Autodesk Platform Services Python client

An unofficial typed Python client and command-line interface for the Autodesk Platform Services (APS) APIs. No affiliation with Autodesk is implied or intended.

APS spans many services behind one host and one OAuth server. This package wraps them in an ergonomic, fully type-hinted client built on Pydantic models, plus an aps CLI for quick access from the terminal.

Features

• One client, every API - thirty-eight APS services mount as namespaces on a single client, from client.auth to client.webhooks, sharing one connection pool and one token cache.
• Typed models - every response is parsed into Pydantic models with descriptive fields.
• Python client and CLI - use it as a library or straight from the shell via aps.
• Sync and async - Client and AsyncClient mount the same services under the same names over httpx.
• Both OAuth flows - 2-legged client credentials and the full 3-legged authorization code flow with PKCE.
• Sensible defaults - caches access tokens per scope set and refreshes them before they expire.
• The awkward parts handled - multipart uploads to signed S3 URLs, translation and workitem polling, URN encoding, JWT assertion signing, cursor paging, and every vocabulary APS uses to report an error.

Services

Each APS API is mounted on the client as its own namespace.

Namespace	API	Reference
client.auth	Authentication (OAuth) v2	docs
client.service_accounts	Secure Service Account v1	docs
client.oss	Object Storage Service v2	docs
client.model_derivative	Model Derivative v2	docs
client.design_automation	Design Automation v3	docs
client.data_management	Data Management v2	docs
client.account_admin	ACC Account Admin v1	docs
client.issues	ACC Issues v1	docs
client.rfis	ACC RFIs v3	docs
client.model_properties	ACC Model Properties v2	docs
client.cost_management	ACC Cost Management v1	docs
client.submittals	ACC Submittals v2	docs
client.sheets	ACC Sheets v1	docs
client.model_coordination	ACC Model Coordination v3	docs
client.data_connector	ACC Data Connector v1	docs
client.takeoff	ACC Takeoff v1	docs
client.assets	ACC Assets v1/v2	docs
client.forms	ACC Forms v1/v2/v3	docs
client.reviews	ACC Reviews v1	docs
client.files	ACC Files v1	docs
client.locations	ACC Locations v2	docs
client.photos	ACC Photos v1	docs
client.relationships	ACC Relationships v2	docs
client.transmittals	ACC Transmittals v1	docs
client.autospecs	ACC AutoSpecs v1	docs
client.classifications	ACC Classifications v1	docs
client.parameters	Parameters v1	docs
client.tandem	Tandem Data v1	docs
client.applications	Application Management v1	docs
client.informed_design	Informed Design v1 (beta)	docs
client.sustainability	Sustainability Data v3 (beta)	docs
client.token_flex	Token Flex Usage Data v1	docs
client.insights	Business Success Plan Reporting v1	docs
client.forma	Forma Site Design v1alpha (alpha)	docs
client.flow_graph	Flow Graph Engine v1	docs
client.building_connected	BuildingConnected v2/v3	docs
client.tradetapp	TradeTapp v2	docs
client.webhooks	Webhooks v1	docs

Installation

pip install autodesk-platform-sdk
# or, with uv:
uv add autodesk-platform-sdk

Requires Python 3.12+.

Configuration

Credentials are read from environment variables (or can be passed directly to Client):

Variable	Description
APS_CLIENT_ID	The application's Client ID.
APS_CLIENT_SECRET	The application's Client Secret.
APS_CALLBACK_URL	A registered Callback URL, used by the 3-legged flow.
APS_ACCESS_TOKEN	A 3-legged access token to act as a user with.
APS_REFRESH_TOKEN	A 3-legged refresh token, exchanged as needed.
APS_BASE_URL	Overrides the APS API host.

export APS_CLIENT_ID="your-client-id"
export APS_CLIENT_SECRET="your-client-secret"
export APS_CALLBACK_URL="http://localhost:53682/callback"

Create an app and its credentials at aps.autodesk.com/myapps. The client credentials grant needs an app registered as Server-to-Server or Traditional Web App; the 3-legged flow additionally needs the callback URL registered on that app, matched exactly - a trailing slash is a different URL.

Timeouts and retries

Every request carries a timeout (default (5, 30) seconds for connect and read) so a stalled connection can't hang the caller forever. Pass timeout= to override it (a single float, a (connect, read) tuple, or None to disable), and retries= to retry connection-establishment failures:

client = Client(timeout=60, retries=3)

retries retries only the connection stage, before any bytes reach the server, so a token is never minted twice. The CLI takes both from the environment instead.

Quick start

The application acts as itself

The 2-legged grant covers most of the platform. Tokens are cached per scope set and reused until they near expiry, so ask for one whenever you need it.

from autodesk_platform_sdk import Client
from autodesk_platform_sdk.services.oss.schemas import PolicyKey

client = Client()

bucket = client.oss.create_bucket("my-app-bucket", PolicyKey.transient)
client.oss.upload_object(bucket.bucket_key, "model.rvt", data)

Acting as a user

Most of APS is scoped to a person rather than an application, and BuildingConnected and TradeTapp accept nothing else. Install a 3-legged token once and every mounted service acts as that user.

from autodesk_platform_sdk import Scope

token = client.auth.authorize_interactively([Scope.data_read, Scope.openid])
client.auth.use_token(token)

client.data_management.list_hubs()  # would report no hubs without this
client.building_connected.list_projects()  # refuses a 2-legged token outright

For an unattended job, a saved refresh token is enough on its own - it is exchanged on first use and again whenever it expires. Refresh tokens are single-use, so persist the replacement:

client = Client(refresh_token=load())
client.auth.on_refresh = lambda token: save(token.refresh_token)

The awkward parts

Several APS workflows take three or four calls and a rule you have to know. Those are wrapped in one method each, so upload_object runs the whole signed-S3 flow with automatic multipart splitting, and wait_for_translation polls to completion:

import pathlib

from autodesk_platform_sdk.services.modelderivative.schemas import OutputType

uploaded = client.oss.upload_object(
    bucket_key, "tower.rvt", pathlib.Path("tower.rvt").read_bytes()
)

client.model_derivative.translate(uploaded.object_id, [OutputType.svf2])
manifest = client.model_derivative.wait_for_translation(uploaded.object_id)
assert manifest.is_successful()

AsyncClient mirrors Client method for method - same arguments, same return types, awaited.

Each service's docstring carries what is peculiar to it: which calls need a subscription, where an empty listing means a permissions failure rather than no data, which IDs carry a b. prefix. Read it in your editor, or with help(client.data_management).

CLI

Every service is a command group. Connection settings come from the same APS_* variables.

aps auth login --scope data:read          # 3-legged, opens a browser
aps oss buckets
aps derivative translate <urn> --output svf2
aps bc projects --include-closed

aps --help lists the groups. aps <group> --help lists its commands.

The CLI reads four more variables of its own:

Variable	Description
APS_API_TIMEOUT	Per-request timeout in seconds.
APS_API_RETRIES	How many times to retry a failed connection.
APS_API_HEADERS	Extra request headers, as a JSON object, to reach APS through a proxy.
APS_CLI_DISABLE	Command paths to hide and refuse, comma separated, such as oss.delete-bucket.

API reference

Authentication (client.auth)

Method	Description
get_access_token	Mint or reuse a 2-legged token for a scope set.
get_active_token	The token every service acts with, refreshing if needed.
use_token	Install a 3-legged token for the whole client to act as a user.
clear_token	Discard the installed token, reverting to 2-legged.
get_authorization_url	Build the 3-legged authorization URL. No I/O.
authorize_interactively	Run the whole 3-legged flow via a browser and local listener.
exchange_code	Exchange an authorization code for a 3-legged token.
refresh_access_token	Exchange a refresh token for a new token pair.
introspect_token	Status and metadata for one of this app's tokens.
revoke_token	Revoke one of this app's tokens.
get_keys	The JWKS used to verify token signatures offline.
get_oidc_spec	The OpenID Connect discovery document.
get_user_info	The profile of the user a 3-legged token belongs to.
exchange_jwt_assertion	Exchange a signed assertion for a service account token.
get_logout_url	Build the logout URL. No I/O.

Service accounts (client.service_accounts)

Method	Description
create_account	Create an account this application can act as.
list_accounts	The service accounts this application owns.
get_account	One service account.
set_account_status	Enable or disable an account.
delete_account	Delete an account and all its keys.
create_key	Create a signing key. Returns the private key once.
list_keys	An account's keys, without their private halves.
set_key_status	Enable or disable one key.
delete_key	Delete one key.
build_assertion	Sign a JWT assertion without exchanging it. No I/O.
get_token	Sign an assertion and exchange it for a token.

Object storage (client.oss)

Method	Description
list_buckets	Buckets this application owns, one page at a time.
create_bucket	Create a bucket. The policy is permanent.
get_bucket_details	One bucket's policy, owner, and permissions.
delete_bucket	Delete a bucket and everything in it.
list_objects	Objects in a bucket, one page at a time.
get_object_details	One object's size, hash, and URN.
upload_object	Upload, running the whole signed-S3 flow.
download_object	Resolve the signed URL and fetch the bytes.
get_signed_upload	Signed URLs to upload parts to. Step one of three.
complete_upload	Assemble uploaded parts into an object. Step three.
get_signed_download	A signed URL to hand to a browser.
copy_object	Copy within one bucket.
delete_object	Delete one object.

Model Derivative (client.model_derivative)

Method	Description
get_formats	Which source formats translate into which outputs.
translate	Start a translation job. Returns once accepted, not finished.
get_manifest	Everything generated from a design, and how far it has got.
wait_for_translation	Poll the manifest until translation finishes.
delete_manifest	Delete a design's derivatives. The source is untouched.
get_model_views	The viewables inside a translated design.
get_object_tree	A view's object hierarchy. None while extracting.
get_all_properties	Properties of every object in a view. None while extracting.
get_thumbnail	A design's thumbnail as PNG bytes.
get_derivative_url	A signed URL for one generated derivative file.

Design Automation (client.design_automation)

Method	Description
list_engines	The engines available to run activities on.
get_engine	One engine's product version and retirement date.
get_engine_health	Whether a product's engines are accepting workitems.
get_engine_stats	How long a product's engines are making work wait.
list_activities	The activities this application can run.
create_activity	Create a job definition.
get_activity	One activity, by fully qualified id.
delete_activity	Delete an activity, its versions, and its aliases.
list_activity_aliases	The aliases on one of your activities.
create_activity_alias	Point a new alias at a version.
get_activity_alias	Which version an alias points at.
update_activity_alias	Move an alias to another version. Releases it.
delete_activity_alias	Delete one alias.
list_activity_versions	The versions one of your activities has.
create_activity_version	Add a version, described in full.
get_activity_version	One numbered version.
delete_activity_version	Delete one version.
list_app_bundles	The app bundles this application can load.
create_app_bundle	Create a bundle and reserve somewhere for its zip.
get_app_bundle	One bundle, by qualified id, with a URL to its zip.
delete_app_bundle	Delete a bundle, its versions, and its aliases.
upload_app_bundle	Upload the zip, running the whole signed-S3 form POST.
list_app_bundle_aliases	The aliases on one of your bundles.
create_app_bundle_alias	Point a new alias at a version.
get_app_bundle_alias	Which version an alias points at.
update_app_bundle_alias	Move an alias to another version.
delete_app_bundle_alias	Delete one alias.
list_app_bundle_versions	The versions one of your bundles has.
create_app_bundle_version	Add a version and reserve somewhere for its zip.
get_app_bundle_version	One numbered version.
delete_app_bundle_version	Delete one version.
create_workitem	Start one run of an activity. Returns pending.
wait_for_workitem	Poll a workitem until it finishes.
get_workitem	How far one workitem has got.
get_workitem_statuses	How far several have got, in one call.
create_workitem_batch	Start several runs in one call.
combine_workitems	Run parts in parallel, then one that merges them.
list_workitems	Your workitems since a moment in time.
delete_workitem	Cancel a workitem that has not finished.
get_nickname	The nickname standing in for the client ID.
set_nickname	Claim a nickname. Only while you own nothing.
delete_app_data	Delete every activity, bundle, and nickname you own.
get_service_limits	The quotas this application runs under.
set_service_limits	Lower your own quotas.
list_shares	What you have shared with other applications.

Every input and output is an HTTP URL, so OSS feeds a workitem directly:

from autodesk_platform_sdk.services.designautomation.schemas import Argument, Verb

source = client.oss.get_signed_download(bucket_key, "tower.dwg")
target = client.oss.get_signed_upload(bucket_key, "tower.pdf")

workitem = client.design_automation.create_workitem(
    "AutoCAD.PlotSheetsetToPDF+prod",
    {
        "SheetSet": Argument(url=source.url),
        "Result": Argument(url=target.urls[0], verb=Verb.put),
    },
)
finished = client.design_automation.wait_for_workitem(workitem.id)
assert finished.is_successful(), finished.report_url

client.oss.complete_upload(bucket_key, "tower.pdf", target.upload_key)

A signed upload URL is one part of a multipart upload, so the object only appears once complete_upload runs. On any failed status, report_url is the engine's log and the only place the failure explains itself.

Activities, app bundles, and engines are addressed two ways, and the service refuses one where it wants the other. get_activity and a workitem's activity take the fully qualified owner.Name+alias. Every alias and version call takes the bare name instead. QualifiedId models the first, and both are checked before the request leaves.

Data Management (client.data_management)

Method	Description
list_hubs	The ACC, BIM 360, and Fusion accounts the caller can reach.
get_hub	One hub.
list_projects	The projects in a hub.
get_project	One project.
get_top_folders	The folders at the top of a project. Where a traversal starts.
get_folder	One folder.
get_folder_contents	What is directly inside a folder, plus each file's tip version.
get_folder_parent	The folder one level up.
search_folder	Search a folder and everything under it, recursively.
get_item	One file, without its contents.
get_item_tip	A file's latest version.
get_item_versions	A file's versions, newest first.
get_item_parent	The folder a file lives in.
get_version	One specific version of a file.
get_version_item	The file a version belongs to.
get_version_download_formats	The formats a version can be exported as.
create_storage	Reserve a place in OSS for a file's bytes.
create_folder	Create a subfolder.
create_item	Create a file and its first version from uploaded bytes.
create_version	Add a version to a file that already exists.
create_download	Start exporting a version as a given file type.
get_download_job	Check on an export job.

BuildingConnected (client.building_connected)

Method	Description
get_current_user	Get the user the token belongs to.
get_user	Get one user at your company.
list_users	List the users at your company.
list_offices	List your company's offices.
get_office	Get one of your company's offices.
list_primary_contacts	List the users designated as an office's primary contacts.
list_certificate_types	List the certificate types BuildingConnected recognizes.
list_certificate_agencies	List the agencies that issue certificates.
list_projects	List the BuildingConnected projects you can reach.
get_project	Get one project.
create_project	Create a project.
update_project	Change a project, sending only the fields given.
delete_project	Delete a project.
list_project_costs	List a project's internal cost breakdown.
create_project_costs	Add cost lines to a project.
update_project_costs	Change cost lines on a project.
delete_project_costs	Remove cost lines from a project.
upload_nda	Upload an NDA document, in both steps.
delete_nda	Delete an NDA document.
get_project_nda	Get the NDA required on a project.
sign_project_nda	Sign a project's NDA as the calling user.
list_team_members	List project team members.
get_team_member	Get one project team member.
add_team_member	Add somebody to a project's team.
update_team_member	Change a project team member, sending only the fields given.
remove_team_member	Remove somebody from a project's team.
list_bid_packages	List bid packages.
get_bid_package	Get one bid package.
create_bid_package	Create a bid package on a project.
update_bid_package	Change a bid package, sending only the fields given.
delete_bid_package	Delete a bid package.
publish_bid_packages	Publish a project's bid packages, making them visible to their bidders.
unseal_bid_packages	Unseal a project's bid packages, making their sealed bids readable.
get_bid_package_stats	Get the response counts for one bid package.
get_bid_package_stats_batch	Get the response counts for several bid packages at once.
list_bid_package_activities	List the recorded activity on bid packages.
list_invites	List invites to bid.
get_invite	Get one invite.
update_invite	Change an invite, sending only the fields given.
invite_bidders	Invite people to a bid package, by email address or user ID.
import_emails	Invite bidders to a bid package by email address alone.
remove_invitee	Remove one person from an invite.
get_invite_certificate	Get a certificate file held by an invited company.
list_bids	List bids.
get_bid	Get one bid.
create_bid	Submit a bid against an invite.
delete_bid	Delete a bid.
list_bid_line_items	List a bid's priced line items.
upload_bid_attachment	Upload a file to attach to a bid, in both steps.
get_bid_attachment	Get one of a bid's attachments, and where to download it.
delete_bid_attachment	Delete a bid attachment.
get_bidding_stats	Get how one bidder company has performed across your projects.
get_bidding_stats_batch	Get bidding performance for several companies at once.
list_project_bid_forms	List project bid forms.
get_project_bid_form	Get one project bid form.
create_project_bid_form	Create a project's bid form.
update_project_bid_form	Replace a project bid form's line items.
list_project_bid_form_line_items	List a project bid form's line items.
create_project_bid_form_line_items	Add line items to a project bid form.
update_project_bid_form_line_items	Change line items on a project bid form.
delete_project_bid_form_line_items	Remove line items from a project bid form.
list_scope_specific_bid_forms	List scope-specific bid forms.
get_scope_specific_bid_form	Get one scope-specific bid form.
create_scope_specific_bid_form	Create a bid package's scope-specific bid form.
update_scope_specific_bid_form	Replace a scope-specific bid form's line items.
list_scope_specific_bid_form_line_items	List a scope-specific bid form's line items.
create_scope_specific_bid_form_line_items	Add line items to a scope-specific bid form.
update_scope_specific_bid_form_line_items	Change line items on a scope-specific bid form.
delete_scope_specific_bid_form_line_items	Remove line items from a scope-specific bid form.
list_opportunities	List your Bid Board opportunities.
get_opportunity	Get one opportunity.
create_opportunity	Create an opportunity on your Bid Board.
update_opportunity	Change an opportunity, sending only the fields given.
delete_opportunity	Delete an opportunity.
list_opportunity_comments	List the comments on an opportunity.
list_opportunity_project_pairs	List the links between opportunities and projects.
get_opportunity_project_pair	Get one opportunity-project pair.
create_opportunity_project_pair	Link an opportunity to a project.
update_opportunity_project_pair	Change which opportunity or project a pair links.
list_contacts	List your company's trade partner and client contacts.
get_contact	Get one contact.
list_preferred_contacts	List the people your offices prefer to deal with at bidder offices.
get_contact_certificate	Get a certificate file held by a contact.

TradeTapp (client.tradetapp)

Method	Description
get_current_user	Get the user the token belongs to, and their company.
list_qualifications	List your subcontractors' submitted questionnaires.
get_qualification	Get one subcontractor's questionnaire in full.
list_office_addresses	List a subcontractor's office addresses.
list_custom_questions	List a subcontractor's answers to your custom questions.
list_financials	List your subcontractors' financial and risk data.
get_financial	Get one subcontractor's financial and risk data in full.
list_flags	List the flags raised against your subcontractors.
get_flag	Get one flag.
create_flag	Raise a flag against a subcontractor.
update_flag	Change a flag, sending only the fields given.
delete_flag	Delete a flag.
list_flag_state_history	List every state a flag has passed through.

Webhooks (client.webhooks)

Method	Description
list_hooks	Every webhook the calling token can see.
list_app_hooks	Every webhook this application owns, whoever created it.
list_system_hooks	The webhooks for one APS service.
list_event_hooks	The webhooks for one event type, optionally by scope.
get_hook	One webhook.
create_hook	Subscribe a callback URL to one event type.
create_system_hooks	Subscribe to every event in a service at once.
update_hook	Change a webhook's status, filter, or attributes.
delete_hook	Delete a webhook.
create_token	Set the application-wide notification secret.
update_token	Replace the notification secret.
delete_token	Remove the notification secret.

ACC Account Admin (client.account_admin)

Method	Description
list_projects	An account's projects, filtered and paged.
get_project	One project. IDs here have no b. prefix.
create_project	Create a project, optionally cloning a template.
list_project_users	A project's members.
get_project_user	One project membership.
assign_project_user	Add someone to a project by email.
update_project_user	Change a member's company, roles, or products.
remove_project_user	Remove someone from a project.
list_companies	An account's companies.
get_company	One company.
list_project_companies	The companies on one project.
list_account_users	The people in an account's directory.
get_account_user	One person from the directory.
search_account_users	Search the directory by name, email, or company.
list_user_projects	The projects one person is on.
list_user_products	The products one person has access to.
list_user_roles	The roles one person holds, and where.
get_business_units	An account's business unit hierarchy.

ACC Issues (client.issues)

Method	Description
get_permissions	What the calling user may do with a project's issues.
list_issue_types	A project's issue types and their subtypes.
list_root_cause_categories	The root causes an issue can be attributed to.
list_attribute_definitions	The custom attributes defined for issues.
list_attribute_mappings	Which issue types each custom attribute applies to.
list_issues	A project's issues, filtered and paged.
get_issue	One issue in full, including what the caller may change.
create_issue	Create an issue against a subtype.
update_issue	Change an issue, sending only the fields given.
list_comments	An issue's comments.
create_comment	Add a comment to an issue.
list_attachments	An issue's attachments.
delete_attachment	Remove one attachment.

ACC RFIs (client.rfis)

Takes the bare project UUID, as Account Admin, Issues, and Submittals do, while Data Management returns the same project as b.<uuid>. Every method strips the prefix.

Listing is search_rfis, a POST: paging, ordering, and filtering travel in the body. Its date filters take a range of two second-precision timestamps joined by .., and refuse a single one.

All 16 endpoints are implemented.

Method	Description
get_permissions	Who the caller is on a project and what its RFIs let them do.
get_workflow	Which review path a project follows, and who may take each part.
search_rfis	A project's RFIs, filtered, ordered, and paged.
get_rfi	One RFI in full, with its responses and the moves open to the caller.
create_rfi	Raise an RFI in one of the statuses open to the caller.
update_rfi	Change an RFI, move it through the workflow, and publish the official response that closes it.
create_response	Answer an RFI as one of its reviewers, and update_response rewrite that answer.
list_comments	The remarks on an RFI, and create_comment add one.
list_attachments	The files on an RFI and on its responses.
upload_attachment	Upload a local file into an RFI's folder, ready to attach to a response.
list_rfi_types	The kinds of RFI a project recognises, with the defaults for each.
list_custom_attributes	The fields a project adds to its RFIs, with create_custom_attribute and update_custom_attribute.
get_custom_identifier	The last RFI number used and the next one free.

ACC Model Properties (client.model_properties)

Indexes the BIM properties of the models on a project and answers queries across all of them. client.model_derivative reads one design at a time; this reads many.

Nothing is indexed until it is asked for. batch_index_status names the file versions wanted and starts the missing indexes as a side effect, so the first call both requests and reports. An index is cached for 30 days from its last use, and the same versions asked for twice give back the same index.

Queries are JSON documents in Autodesk's own query language, passed as dictionaries. An index row keys its properties by field key, so get_index_fields is what turns p153cb174 back into name.

Takes the bare project UUID, as Account Admin, Issues, RFIs, and Submittals do, while Data Management returns the same project as b.<uuid>. Every method strips the prefix. A 2-legged token is refused outright, where Submittals and RFIs answer one the same way they answer any caller.

All 16 endpoints are implemented, plus four helpers over them.

Method	Description
batch_index_status	The index of each file version, building the ones that are missing.
get_index	How far one index has got, and where its resources live.
wait_for_index	Poll an index until it stops building.
get_index_manifest	Which files and viewables an index was built from.
get_index_fields	The columns an index holds, and the key each one is stored under.
get_index_properties	Every row of an index.
create_index_query	Start a query over an index, with get_index_query for its progress.
get_index_query_properties	The rows one query matched.
run_index_query	Start a query, wait for it, and return the rows, in one call.
batch_diff_status	The diff of each version pair, building the ones that are missing.
get_diff	How far one diff has got, and how much it found changed.
wait_for_diff	Poll a diff until it stops building.
get_diff_manifest	Which pair of file versions a diff compared.
get_diff_fields	The columns a diff holds.
get_diff_properties	Every added, removed, and changed design element.
create_diff_query	Start a query over a diff, with get_diff_query for its progress.
get_diff_query_properties	The changes one query matched.
run_diff_query	Start a query, wait for it, and return the changes, in one call.

ACC Cost Management (client.cost_management)

Keyed by container ID, not project ID. On an ACC project the container is the project's own UUID, and every method strips Data Management's b. prefix. On a BIM 360 project the two can differ, so read the right one with client.data_management.get_container_id(hub_id, project_id, Container.cost).

All 98 endpoints are implemented. The table names the entry point of each group; each has the get, create, update, and delete methods its resource supports.

Method	Description
list_budgets	The owner-approved budget lines, with their rolled-up amounts.
list_templates	The budget code templates, and list_segments their grammar.
list_values	The allowed values of one budget code segment.
list_contracts	The supplier contracts, the commitment side of the picture.
link_budgets	Link budgets to contracts, and unlink others, in one call.
list_main_contracts	The prime contracts, and list_main_contract_items their lines.
list_change_order_workflows	Which of the five change order types the project runs.
list_change_orders	The change orders of one type.
list_cost_items	The priced lines of a change order.
list_sub_cost_items	One cost item's breakdown, per estimate stage.
list_schedule_of_values	A contract's schedule of values.
list_expenses	Costs recorded outside a contract, and their items.
list_payments	The payment applications, and list_payment_items their lines.
list_attachments	The files attached to cost items, and list_documents the generated ones.
list_properties	The custom attributes defined for a kind of cost item.
list_taxes	The tax formulas applied to a cost object.
list_performance_tracking_items	The budgets enrolled in performance tracking.
list_time_sheets	Quantities tracked against them over a period.
list_actions	What one item can do next, and list_action_histories what it has done.

ACC Submittals (client.submittals)

Takes the bare project UUID, as Account Admin and Issues do, while Data Management returns the same project as b.<uuid>. Submittals answers a prefixed ID with a 500 rather than a readable error, so every method strips the prefix.

All 33 endpoints are implemented.

Method	Description
get_permissions	What the calling user may do with a project's submittals.
get_metadata	Every vocabulary a project has configured, in one call.
list_items	A project's submittal items, filtered and paged.
get_item	One item in full, including the workflow moves open to the caller.
create_item	Create an item against a type, a spec section, and a start state.
update_item	Change an item, sending only the fields given.
transition_item	Move an item to another state in the review workflow.
list_revisions	The past rounds of submission an item has been through.
get_next_custom_identifier	The next number free to assign, and validate_custom_identifier whether one is.
change_sequence_type	Switch a project between global and per-spec numbering. Beta.
list_attachments	The files on an item.
upload_attachment	Attach a local file, running the reserve, upload, and complete calls.
create_attachment	Link a file version that is already in the Files tool.
update_attachment	Mark a local upload complete.
list_steps	The rounds of review an item goes through, and get_step one of them.
list_tasks	The reviews one step waits on, and get_task one of them.
close_task	Close a review task by giving its response.
list_packages	The packages items are grouped into, and get_package one of them.
list_responses	The verdicts a reviewer can pick, and get_response one of them.
list_specs	A project's spec sections, with get_spec and create_spec.
list_item_types	The kinds of submittal a project recognises, and get_item_type one.
list_templates	The review templates an item can be created against.
list_mappings	Who holds the submittal manager role, with create_mapping and delete_mapping. Both writes are beta.
get_async_job	The progress of work Submittals runs in the background. Beta.

ACC Sheets (client.sheets)

Holds the published drawing sets of a project: the version sets that date each issue, the sheets inside them, and the upload that produces them.

Takes the bare project UUID. This API accepts Data Management's b.<uuid> form too, where Submittals, RFIs, and Model Properties all refuse it, but every method strips the prefix anyway.

Getting a drawing in is a five-call flow across two APIs. upload_sheets runs the first three: reserve OSS storage, put the bytes there, and start the extraction. Sheets then splits the file into review sheets, reading a number and a title off each title block by OCR. Correct them with update_review_sheets, then publish_review_sheets turns each into a sheet. Nothing reaches the project until that last call.

A sheet number is unique within a version set, not within a project, so the same number appears once per issue. list_sheets returns only the copy in the latest-dated version set unless asked otherwise.

All 22 endpoints are implemented.

Method	Description
list_version_sets	The dated issues a project's sheets are grouped into, and get_version_sets up to 200 by ID.
create_version_set	Open a new dated issue, and update_version_set rename or redate one.
delete_version_sets	Delete up to 10 version sets, and every sheet inside them.
upload_sheets	Put local PDFs into a version set, running the reserve, upload, and extract calls.
create_storage	Reserve the OSS object a file's bytes go into, and create_upload start the extraction.
list_uploads	A project's uploads, and get_upload one of them.
wait_for_upload	Poll an upload until it reaches review, publication, or failure.
list_review_sheets	The pages an upload pulled out of its files, before publishing.
update_review_sheets	Correct the number, title, or tags OCR read off the title blocks.
get_thumbnails	Preview images of up to 100 review sheets at once.
publish_review_sheets	Turn an upload's review sheets into sheets in its version set.
list_sheets	A project's published sheets, filtered and paged, and get_sheets up to 200 by ID.
update_sheets	Renumber, retag, or reissue up to 200 sheets in one call.
delete_sheets	Delete up to 200 sheets, and restore_sheets put them back.
create_export	Render up to 1000 sheets into one PDF, and wait_for_export poll for its link.
get_export	How far an export has got, and its download link once it is done.
list_collections	The collections a project's sheets can be grouped into, and get_collection one.

ACC Model Coordination (client.model_coordination)

Holds the coordination spaces a project clash tests against, and the results of those tests.

Takes the container ID, not the project ID. On an ACC project the container is the project's own UUID: the web app allocates it under the Docs project ID, so no relationship lookup is needed and every method strips a b. prefix anyway. A project whose administrator has not yet configured a coordination space has no container at all, and answers 404 whatever the caller does.

Two services sit behind one namespace. Model sets, versions, and views come from /bim360/modelset/v3. Clash tests and clash groups come from /bim360/clash/v3.

Nothing here runs a clash test. The service takes a model set version whenever the documents under the model set's folder change, then clash tests it on its own. So the read path is get_latest_model_set_version, then list_version_clash_tests, then list_clash_test_resources and download_clash_resource.

Clash results are files rather than JSON: three gzipped documents holding the clashing object pairs, the Viewer objects behind them, and the documents those objects live in. Their links expire, so ask for them again rather than storing one.

Every write is asynchronous and answers with a job. wait_for_container_job polls the one a new model set starts, and wait_for_model_set_job polls every other model set write.

All 43 endpoints are implemented. The screenshot pair is documented twice, on both service roots, and is written once.

Method	Description
list_model_sets	The coordination spaces on a project, and get_model_set one with its folder and tip version.
create_model_set	Open a coordination space over one Docs folder, and update_model_set rename or hide it.
list_model_set_versions	The versions a model set has captured, with get_model_set_version and get_latest_model_set_version.
create_model_set_version	Sample the folder now, for when automatic versioning is off.
enable_model_set_versions	Resume versioning on every Docs change, and disable_model_set_versions stop it.
list_views	The saved views on a model set, with get_view, create_view, update_view, and delete_view.
search_view_lineages	The views tracking a given set of document lineages. Deprecated by Autodesk.
list_view_versions	Which models every view resolved to at one model set version, and get_view_version for one.
list_clash_tests	The clash tests run on a model set, and list_version_clash_tests those on one version.
get_clash_test	One clash test and whether its results are ready.
list_clash_test_resources	The signed links to a finished test's three result files.
download_clash_resource	Fetch one result file and gunzip it.
close_clash_groups	Close batches of clashes needing no work, and reopen_clash_groups put them back.
assign_clash_groups	Raise batches of clashes as issues for somebody to fix.
search_closed_clash_groups	A model set's closed groups, and search_assigned_clash_groups its assigned ones.
list_test_closed_clash_groups	Replay closed groups against a later test, splitting each into still-clashing and resolved.
get_closed_clash_groups	Up to 20 groups with the data needed to draw them, and get_assigned_clash_groups by group or issue ID.
get_grouped_clashes	Every clash already closed or assigned, to subtract from a test's results.
add_model_set_issues	Raise inspection issues pinned to points in the coordinated model.
get_issue_view_context	The documents to load to reopen an issue, and get_assigned_clash_group_view_context for a clash issue.
add_screenshot	Upload a PNG for an issue, a clash group, or a view to claim, and get_screenshot read one back.
get_container_job	The job a new model set started, with get_model_set_job, get_view_job, and get_clash_group_job.
wait_for_container_job	Poll a job until it settles, and wait_for_model_set_job for every other model set write.

ACC Takeoff (client.takeoff)

Reads the quantities a project has taken off its drawings and models, package by package.

Project keyed rather than container keyed, so there is no relationship lookup. The reference says a b. prefix works, and it does, but every method strips one anyway so that one form reaches the wire.

Almost all of it is read only. Settings, packages, and the assigned classification structures accept a write. Takeoff types and takeoff items do not, because only the web app can measure something.

A takeoff type is the template and a takeoff item is one measurement made against it. Classifications live on the type, so an item is classified by reading the type its takeoff_type_id names.

The six classification system endpoints are deprecated and are not implemented. A migrated project answers them 409 Conflict and keeps its classifications in the Classifications API, reached from here through list_assigned_structures.

All 14 live endpoints are implemented.

Method	Description
get_settings	The units a project measures in, and update_settings to change them.
list_packages	The takeoff packages a project holds, and get_package one.
create_package	Open a new package, and update_package rename one.
list_takeoff_types	The measurable things defined in a package, and get_takeoff_type one.
list_takeoff_items	The measurements made in a package, and get_takeoff_item one.
list_content_views	The sheets and model views a project can be taken off against.
list_assigned_structures	The classification trees assigned to a project's takeoffs.
add_assigned_structures	Assign up to five trees, and remove_assigned_structure unassign one.

ACC Assets (client.assets)

Tracks the physical equipment on a project, and the categories, statuses, and custom attributes that describe it.

Project keyed rather than container keyed, so there is no relationship lookup. The reference says a b. prefix works, and it does, but every method strips one anyway so that one form reaches the wire.

The API is versioned per resource: the asset endpoints are on v2 and everything else is still on v1. Both roots take the same project ID.

Almost every write is a batch. An asset is created, changed, and deleted only through create_assets, update_assets, and delete_assets, each of which takes up to 100 assets and is all or nothing. Deletion is soft everywhere, so a deleted record keeps its ID and comes back to any read that passes include_deleted.

A category decides both the statuses an asset may hold and the custom attributes it may carry, and a child category inherits both. An asset's custom_attributes are keyed by an attribute's name, such as ca1, so list_custom_attributes is the lookup table.

list_error_codes and get_error_code need no token and no project. They turn the errorCode on a failure into a reason a caller can act on.

All 24 endpoints are implemented.

Method	Description
list_assets	A project's assets, filtered and paged by cursor, and get_assets up to 100 by ID.
create_assets	Create up to 100 assets, update_assets patch them by ID, and delete_assets remove them.
list_categories	A project's whole category tree, and get_categories some of it by ID.
create_category	Add a category under an existing one.
list_status_sets	The status sets a project defines, each with its statuses, and get_status_sets by ID.
create_status_set	Create a status set and its statuses in one call.
get_category_status_sets	Which status set each category draws from, and assign_status_set to change one.
list_statuses	Every status across every set, and get_statuses some by ID.
create_status	Add a status to an existing set.
list_custom_attributes	Every custom attribute a project defines, and get_custom_attributes by ID or name.
create_custom_attribute	Define an attribute, and update_custom_attribute change one.
list_category_custom_attributes	What one category offers, and assign_custom_attribute to add to it.
list_error_codes	The Assets error catalogue, and get_error_code one entry.

ACC Forms (client.forms)

Reads the daily logs, checklists, and inspection records filled in on site, and the templates they are made from.

Project keyed rather than container keyed, so there is no relationship lookup. Every method strips the b. prefix Data Management adds.

The API is versioned per endpoint and three roots run side by side. The 2026 April release put a new forms listing, new value writes, and the section reader on v2, left their deprecated twins and everything else on v1, and gave weather a v3 root of its own.

Two calls therefore exist twice. list_forms and update_values are the deprecated v1 pair, which Autodesk removes on 2026-12-29; list_forms_v2 and update_values_v2 replace them and are in public beta. The two answer differently shaped records - Form against FormV2 - and different status vocabularies, so migrating is a real change rather than a rename. Both are here so a caller can move at their own pace.

A template defines the questions and a form is one instance of it. Read a native template's structure with get_layout and then get_section per section, which is the only place a custom table's column IDs are published. A PDF template answers through pdf_values instead; the API can neither create a template nor replace a form's PDF.

Values are typed by the template rather than by the SDK, so an answer carries one of text_val, number_val, toggle_val and the rest, and value_name says which. Writing is a merge, not a replace: a call touches only the fields it names, and at most 10 of them, so a long form is filled in over several calls.

All 13 endpoints are implemented.

Method	Description
list_templates	The form templates a project defines, with their permissions.
get_layout	A template's structure and its sections, and get_section one section in full.
list_forms	A project's forms with their answers inline. Deprecated; removed 2026-12-29.
list_forms_v2	A project's forms, with include deciding how much of each one comes back.
create_form	Create a form from a template, and update_form change it or move it through its workflow.
update_values	Fill in fields and built-in table rows. Deprecated; removed 2026-12-29.
update_values_v2	Fill in fields and table rows, custom tables included.
delete_values	Remove table rows from a form.
list_values	A form's non-tabular answers, paged, and list_table_values one table's rows.
get_weather	The weather captured for a form's date.

ACC Reviews (client.reviews)

Reads the document approval workflows that gate a file version's status in ACC Docs, and the reviews running on them.

Project keyed rather than container keyed, so there is no relationship lookup. The API takes the project UUID with or without the b. prefix, and every method strips the prefix so one form reaches the wire.

The API is read mostly. create_workflow and create_review are the only writes. Nothing here approves, rejects, claims, voids, or advances a review, and nothing here edits or deletes a workflow. Those happen in the ACC UI, and this API watches them.

A workflow is the template and a review is one run of it over a set of file versions. get_review_workflow reads the snapshot the review captured when it started, which is why it can differ from get_workflow on the same ID.

list_version_approval_statuses is the bridge to client.data_management: it takes a Data Management version URN and answers what every review decided about it. A version reads IN_REVIEW while any review holding it is still open.

Every method takes user_id, which sends x-user-id. Autodesk requires it on a 2-legged call, since a review is always somebody's.

All 10 endpoints are implemented.

Method	Description
list_workflows	A project's approval workflows, active ones unless asked otherwise.
get_workflow	One approval workflow as it stands now.
create_workflow	Create an approval workflow. It cannot be edited or deleted afterwards.
list_reviews	A project's reviews, narrowed by status, workflow, dates, or who owes the next step.
create_review	Start a review over up to 1000 file versions.
get_review	One review, its status, and who owes its current step.
get_review_workflow	The workflow snapshot a review captured when it started.
list_review_progress	What happened at each step of a review, newest first.
list_review_versions	The file versions in the current round of a review.
list_version_approval_statuses	What every review decided about one file version.

ACC Files (client.files)

Covers the parts of the ACC Files tool that client.data_management does not reach: PDF export, folder permissions, custom attributes on documents, file naming standards, file packages, and the Revit models linked into a published model.

This is one documentation tab over four services. The endpoints sit on four roots: construction/files/v1 for export and project custom attributes, bim360/docs/v1 for permissions, folder attributes, and naming standards, construction/packages/v1 for packages, and construction/rcm/v1 for linked Revit files. All four take the project UUID with or without the b. prefix, and every method strips the prefix so one form reaches the wire.

Nothing here lists folders, items, or versions. Data Management owns those. get_versions is the one deliberate overlap: it reads up to 50 versions in one call and adds the approval status and the custom attribute values that JSON:API leaves out. There is no upload here either, because a file reaches the Files tool through client.data_management and client.oss.

Seven endpoints are beta: the six custom attribute calls and the naming standard.

Every method except get_naming_standard takes user_id, which sends x-user-id. The two export methods need it on a 2-legged call, though their reference calls user context optional; without it the wire answers 400 ERR_BAD_INPUT, User ID is required. get_naming_standard takes a region instead, and it is the only endpoint here that does.

All 16 endpoints are implemented, plus wait_for_export and download_export.

Method	Description
export_files	Start a PDF export of up to 200 file versions.
get_export	Where a PDF export job has got to, and its signed download link.
wait_for_export	Poll an export job until it stops, whatever the outcome.
download_export	Resolve the signed link and fetch the exported PDF or ZIP.
list_folder_permissions	What every user, role, and company may do in one folder.
create_folder_permissions	Grant folder permissions to subjects that have none.
update_folder_permissions	Replace the permissions a subject holds. They are not added to.
delete_folder_permissions	Remove every direct permission a subject holds on a folder.
get_versions	Up to 50 file versions with their approval status and custom attributes.
list_folder_attribute_definitions	The custom attribute definitions one folder carries.
create_folder_attribute_definition	Add a custom attribute definition to a folder.
list_custom_attribute_definitions	The custom attribute definitions a whole project carries.
list_custom_attribute_items	The options of a large drop-down custom attribute.
update_version_attributes	Write custom attribute values onto a file version. A null clears one.
get_naming_standard	One file naming standard and the format it enforces.
list_packages	A project's file packages, narrowed by creator, date, or version type.
list_package_resources	The file versions one package holds, with their approval status.
list_linked_files	The Revit models linked into a published model, each with a signed URL.

ACC Locations (client.locations)

Reads and edits the location breakdown structure a project tags its work against: a tree of building areas, up to 20 tiers deep and 10,000 nodes in all.

This is the lookup for the location IDs the other services return. Issues, assets, forms, photos, RFIs, and submittals each carry a node ID. Nothing else turns one into a name. list_nodes(project_id, node_ids=[...]) answers with the node and its ancestors.

Every path is tree scoped as well as project scoped. Autodesk accepts the literal default as the tree ID and nothing else, since a project holds one tree, so tree_id defaults to it.

Locations refuses the b. prefix Data Management adds, with a 400, so every method strips one. Autodesk does not serve BIM 360 projects here.

The root node is created with the project and cannot be created or deleted. Read its ID off list_nodes, where it is the one node with no parent.

delete_node deletes the whole subtree below the node as well, and everything tagged against any of those nodes loses its location.

All 4 endpoints are implemented.

Method	Description
list_nodes	A project's location nodes, all of them or the ones named.
create_node	Add a node under a parent, last among its siblings or beside a named one.
update_node	Rename a node, rebarcode it, or both.
delete_node	Delete a node and every node below it.

ACC Photos (client.photos)

Reads the photos and videos captured against a project: stills, infrared, photospheres, and video, each with where and when it was taken, who added it, and which tool it came from.

Nothing here writes. Photos publishes a read of one record and a filtered listing of many, and media is added through the Photos tool itself.

Both endpoints need a 3-legged token. An application token answers 401 Authorization failed on every path.

The listing is a POST, and its paging goes in the body rather than the query string. Read the next cursor off Page.next_cursor() and pass it back as cursor_state, which makes Autodesk ignore every other argument.

Asking for PhotoInclude.signed_urls adds storage links to the image bytes. They carry their own credentials and expire in about a minute, so fetch with them and never store them. download_photo does the read and the fetch in one call.

Both reference pages show a curl URL with no v1 segment. That path is a gateway 404; the one in the "Method and URI" row is the one that answers.

All 2 endpoints are implemented.

Method	Description
get_photo	One photo, optionally with signed links to its bytes.
filter_photos	A project's photos, narrowed by ID, author, kind, title, or date.
download_photo	Resolve a photo's signed link and fetch the bytes.

ACC Relationships (client.relationships)

Reads and writes the links that join a record in one ACC service to a record in another: an asset to a document, an RFI to a sheet, a form to a photo.

A relationship is an ID joining two entities, and each entity is a domain, a type, and an ID. The link is read in either direction and holds no other data. Holding one grants no access to either record, because the owning service still applies its own permissions when the record is read.

Every path but one takes a container ID. The relationships container is created with the ACC Docs project and carries that project's own ID. So the container is the bare project UUID, and no lookup is needed. Autodesk refuses Data Management's b. prefix with a 400, so every method strips it.

Every endpoint needs a 3-legged token, including get_writable_domains, which takes no container. An application token answers 403 on every path.

Call get_writable_domains before writing. It reports which domain and entity type pairs the caller may link, and it is the only complete record of what the service supports: a live answer names eight domains the querying tutorial's tables leave out.

Two calls are not what their HTTP verb suggests. delete_relationships is a POST that deletes, and intersect_relationships is a POST that reads.

delete_relationships cannot be undone, and a deleted relationship leaves no trace in either record it joined.

Autodesk's own examples for sync_relationships send an empty body, which the service refuses with "At least one parameter SyncToken or Domains need to be completed". SyncRequest matches the service and refuses it locally.

All 9 endpoints are implemented.

Method	Description
get_writable_domains	The domain and entity type pairs this user may link.
add_relationships	Link up to 20 pairs of records in one call.
delete_relationships	Delete up to 50 relationships by ID. Cannot be undone.
get_relationship	One relationship by ID.
get_relationships	Up to 50 relationships by ID.
search_relationships	Relationships matching one entity, or a pair of them.
intersect_relationships	What a batch of up to 20 known records is related to.
sync_relationships	Replicate a container's relationships, or the changes since a token.
get_sync_status	Whether up to 3 sync tokens have anything waiting behind them.

ACC Transmittals (client.transmittals)

Reads the transmittals a project has issued: the formal handover of a set of documents to a set of recipients, with a cover note and a tracked receipt.

A transmittal is a snapshot. It records the document versions and folders as they stood when it went out, and who it was addressed to. list_recipients adds when each recipient received, viewed, and downloaded it.

Nothing here writes, and nothing here sends. Autodesk publishes five reads and no other verb, so a transmittal is created and issued in the ACC Transmittals tool. The API also cannot change a transmittal's settings, add a recipient, or export a package.

A transmittal takes a moment to package. While Transmittal.status is SENDING, the three sub-reads answer 202 Accepted with nothing in them. All three methods return None to say so, rather than an empty result.

recipients on a transmittal is the addressing that was chosen, by user, company, and role. list_recipients is the people it resolved to, with their receipt timestamps.

The three paged listings sort by three different sets of fields, and each takes its own enum. The direction is always sent, because the pages disagree about what a bare field name means.

All 5 endpoints are implemented.

Method	Description
list_transmittals	The transmittals a project has issued.
get_transmittal	One transmittal, with its sender and addressing.
list_recipients	Everyone the transmittal reached, with receipt timestamps.
list_folders	The folders the transmittal included.
list_documents	The document versions the transmittal included.

ACC AutoSpecs (client.autospecs)

Reads the draft submittal log AutoSpecs extracts from a project's specification.

Somebody uploads a spec book to the project in the ACC AutoSpecs tool. AutoSpecs reads the submittal requirements out of it into a Smart Register. That register is the list of submittals the specification asks for, before anyone turns them into real ACC submittals through client.submittals.

Nothing here writes. Autodesk publishes four reads and no other verb, so a spec book is uploaded, extracted, and edited in the AutoSpecs tool. The API also cannot filter submittals, edit the Smart Register, or reach the Spec View and Product Data tools.

Every read but get_metadata is scoped by a specification version, and get_metadata is where the version IDs come from. A version ID is an AutoSpecs number such as 2268, not a Data Management version URN. The service refuses anything that is not a signed 32-bit integer.

Reading a spec book takes minutes and AutoSpecs publishes no webhook. Autodesk says the Smart Register is available only once SpecVersion.status reads Completed, and the field guide says to poll get_metadata for it. wait_for_extraction does that polling.

Every submittal type, submittal group, and region is a plain string. Autodesk's own examples contradict its own field tables on the case of all three. The Retrieve a Smart Register tutorial goes further and answers a submittal type the reference does not list.

All 4 endpoints are implemented, plus wait_for_extraction.

Method	Description
get_metadata	A project's specification versions and their extraction status.
wait_for_extraction	Poll one version until AutoSpecs stops reading it.
get_smart_register	Every submittal the specification asks for.
get_requirements	Those submittals counted by division, section, and group.
get_submittals_summary	Those submittals counted by group and by type.

ACC Classifications (client.classifications)

Reads and writes the classification trees a project tags its work against: the versioned taxonomies such as Uniformat and MasterFormat.

This is the lookup for the classification IDs Takeoff returns. A migrated Takeoff project reports a structureId and nodeId pair on every quantity definition. Nothing else turns that pair into a code and a name, and the Takeoff structureId is the tree ID here.

Autodesk marks the whole module beta and may change it without a migration guide.

A node cannot be created, renamed, moved, or deleted on its own. Every change to the shape of a tree goes through import_tree or reimport_tree, which publish a whole new version. Autodesk's own Delete a Node and Restore a Node guides are both a reimport.

reimport_tree replaces the tree rather than adding to it. Every node missing from its payload is marked deleted in the new version, so send the whole tree and give each node its existing id. An empty payload deletes every node, so allow_empty guards it and defaults to False.

Only the tip version is readable. list_nodes answers from the latest published version, and Autodesk publishes no way to read an older one. include_deleted adds the nodes an earlier reimport dropped.

Classifications refuses the b. prefix Data Management adds, with a 400, so every method strips one.

All 6 endpoints are implemented.

Method	Description
list_trees	The classification trees in a project.
get_tree	One tree, with where it came from.
update_tree	Rename a tree, redescribe it, or both.
import_tree	Create a tree and publish its first version.
reimport_tree	Replace a tree's nodes and publish a new tip version.
list_nodes	The nodes of a tree's latest published version.

Tandem Data (client.tandem)

Reads and writes the digital twin data Autodesk Tandem holds: facilities, their models, the properties on every element, the documents attached to them, and the live sensor streams.

Tandem calls a facility a twin, and the two words mean the same thing. A facility is built from source models. Every logical element lives in its default model, whose URN is the facility URN with dtm in place of dtt, and default_model_id renders it.

Permission comes from the Tandem product rather than from the OAuth scope. A user or an application reaches nothing until a Tandem administrator adds it to a facility or to a team, whatever scopes its token carries. Add an application by client ID: the User tab of a facility, or the Team tab of the Manage page for a whole account.

Property data speaks qualified property names rather than JSON keys. A name is a column family and a column. n:n is the common name and z:3wc is a user-defined parameter. get_model_schema maps those onto the names a person recognises. They are chosen per model, so a facility of three models needs three scans.

mutate_elements writes them back, pairing a list of element keys with a list of mutations. Two of its four operations destroy rather than change: delete_row deletes the element and, on a stream, every reading it holds, and delete outside family z clears a property the source model owns. allow_destructive guards both and defaults to False.

reset_stream_secrets issues new secrets and stops every sensor still posting with the old ones. Read a secret with get_stream_secrets instead. delete_stream_data destroys readings, so it needs either the substreams to clear or an explicit all_substreams.

Tandem is deployed in US, EMEA, and AUS only. The gateway refuses the other five members of Region with a 400, so region refuses them first.

All 32 endpoints are implemented.

Method	Description
list_groups	The Tandem accounts you can reach.
get_group	One account, with its facilities and members.
get_group_history	What changed across an account.
set_group_user_access	Set what one person may do with an account.
list_twins	The facilities of an account, keyed by URN.
create_twin	Create a facility in an account.
get_twin	One facility, with its models, documents, and template.
check_twin_access	What you may do with a facility, from the headers.
create_default_model	The model that holds logical elements and streams.
get_twin_history	What changed in a facility.
list_twin_users	Who may see a facility.
list_views	The saved views of a facility.
get_inline_template	The facility template, classification and all.
add_document	Attach a document to a facility.
get_document	One document, with a signed link to the file.
delete_document	Remove Tandem's copy of a document.
get_model_schema	The property definitions of one model.
scan_properties	Read properties off the elements of one model.
mutate_elements	Write properties onto the elements of one model.
create_element	Add an element, such as a stream.
get_model_history	What changed in one model.
list_stream_configs	How every stream in a model reads its values.
get_stream_config	How one stream reads its values.
update_stream_configs	Change several stream configurations at once.
save_stream_config	Replace one stream's whole configuration.
get_stream_secrets	The secrets that sign the ingestion URLs.
reset_stream_secrets	Issue new secrets, invalidating the old ones.
get_latest_stream_data	The last known reading of each named stream.
get_stream_data	The readings one stream recorded over a range.
send_stream_data	Record one reading against a stream.
send_stream_webhook	Record readings for several streams at once.
delete_stream_data	Delete readings from named streams.

Application Management (client.applications)

Reads the APS applications you own or collaborate on, the API products they may reach, and what they consumed day by day. It also rotates their client secrets.

commit_secret_rotation retires a live credential and cannot be undone. Rotation runs in two steps. prepare_secret_rotation issues a second secret while the current one keeps working, and commit_secret_rotation makes the new one active and the old one dead. Deploy the new secret everywhere first, with each caller trying it and falling back to the old one, then commit. A caller still on the old secret is locked out the moment the commit lands.

Each rotation call needs a client assertion, which is a JWT the application signs with a key published at its own JWKS URI. Set that URI on https://aps.autodesk.com/myapps before the first rotation.

The reads want a 3-legged token carrying application:client:read, or a Secure Service Account token, which client.service_accounts mints for a headless server. The three rotation calls want the application's own 2-legged token carrying application:client:rotate_secret, so clear any user token first with client.auth.clear_token().

Nothing here registers, edits, or deletes an application. Autodesk publishes eight operations and documents no others, and the application:client:write scope its API Basics page names has no v1 endpoint.

The listings page by URL alone. There is no offset and no cursor, so Page.next_url() reports where the next page lives and the listing ends when that comes back None.

All 8 endpoints are implemented.

Method	Description
list_api_products	The API products an application can be given access to.
list_applications	The APS applications you own or collaborate on.
get_application	One application, without its client secret.
list_collaborators	Who else has access to one application.
get_usage_daily_totals	What one application consumed, day by day.
prepare_secret_rotation	Issue a second secret, leaving the current one live.
commit_secret_rotation	Make the prepared secret active and retire the old one.
cancel_secret_rotation	Discard a prepared secret.

Informed Design (client.informed_design)

Publishes a configurable manufactured product from an Inventor or Fusion model, then generates design outputs from configurations of it. This API is beta, and Autodesk says not to build production software on it yet.

Four resources form a chain, and Autodesk's walkthrough runs them in order. A product holds releases, a release holds variants, and a variant holds outputs. Create the product, upload its design data and rules, create a release over them, create a variant that sets the parameters, then generate outputs.

Every call but the three rules ones needs an access_type and an access_id, which decide what the caller may see. For ACC, BIM360, and FUSION the ID is <projectId>|<folderUrn>, and for USER it is the Autodesk user ID. Autodesk authorizes on this pair as well as on the token, so a wrong one is refused before the resource is looked up. list_upload_requests is the only endpoint that takes USER.

Every call needs a 3-legged token carrying data:read, data:write, data:create, and account:read. Autodesk recommends a Secure Service Account for a headless server, which client.service_accounts mints.

Output generation and output upload are asynchronous. Both answer a request whose status starts at PENDING, and wait_for_outputs and wait_for_upload poll until it settles. Variant creation is not asynchronous, whatever the walkthrough's ordering suggests.

update_release replaces the whole output settings list, so an empty one leaves the release able to generate nothing. allow_empty defaults to False and has to say so on purpose.

All 31 endpoints are implemented.

Method	Description
list_products	The configurable products in one container.
create_product	Create a configurable product.
get_product	One configurable product.
update_product	Change a product's name, state, default release, or description.
delete_product	Delete a product and everything under it.
get_upload_urls	Signed links for uploading one file in parts.
complete_upload	Mark every part of a file as arrived.
get_download_url	A signed link to one uploaded file.
delete_upload	Delete one uploaded file.
download_product_file	Sign a link and fetch the bytes behind it.
list_releases	The releases of one product.
create_release	Create a release from an uploaded design data set.
get_release	One release, with its parameters and output settings.
update_release	Change a release's state, notes, or output settings.
delete_release	Delete a release and every variant under it.
list_variants	The variants of one release.
create_variant	Create a variant by setting a release's parameters.
get_variant	One variant and the values it was configured with.
delete_variant	Delete a variant and every output from it.
list_outputs	Generated outputs, narrowed by at least one filter.
create_outputs	Start generating outputs from one variant.
get_output	One generated output.
delete_output	Delete one generated output.
get_outputs_request	One generation run and how far it has got.
wait_for_outputs	Poll a generation run until it settles.
list_upload_requests	The runs that copied outputs into a folder.
create_upload_request	Start copying outputs into an ACC, BIM 360, or Fusion folder.
get_upload_request	One upload run and how far it has got.
wait_for_upload	Poll an upload run until it settles.
create_download_request	Signed links to a set of generated outputs.
get_download_request	One download request again, with fresh links.
evaluate_rules	Run a rules document over a set of parameter values.
validate_rules	Check whether a rules document is well formed.
get_rules	Read a rules document.

Sustainability Data (client.sustainability)

Reads embodied carbon and other environmental impact figures for construction materials, through one interface over several third-party providers. This API is beta and Autodesk says its paths may change at general release.

Start with list_datasets. A dataset is one provider's collection, and its supported_data_types say which of the five resources it serves: baselines, generic estimates, product EPDs, industry EPDs, and activities. Calling a resource a dataset does not serve answers 424 Failed Dependency rather than an empty page.

Every resource offers the same shapes. A filter_* call lists records, a get_* call reads one, and a filter_*_impacts call reads its life cycle impact figures. Product EPDs and industry EPDs add a statistics summary, and activities add exchanges instead of categories.

Read get_*_filters before building a filter. Filter support varies by dataset. A filter the dataset does not know is dropped rather than refused, so check ignored_filters on the page that comes back. A dataset can also narrow a filter it does accept, which it reports in applied_filters.

from autodesk_platform_sdk.services.sustainability.schemas import FilterClause

page = client.sustainability.filter_product_epds(
    dataset_id,
    filters={"jurisdictions": FilterClause(op="IN", value=["US"])},
    limit=100,
)

Nothing here writes. Every listing is a POST because the filter goes in the body, and paging is limit and offset in the query.

Every call needs a 3-legged token carrying data:read, and the application needs the Sustainability Data API added to it on https://aps.autodesk.com/myapps. A client ID without that answers 403 AUTH-001 on every path. Each provider also grants a separate data licence, so a dataset appears in list_datasets only once yours is in place.

The base path is split. The five construction resources are served under /sustainability/v3beta and activities and flows under /sustainability/v3. Autodesk's Quick Reference page writes all seven activity and flow routes under v3beta, and that prefix is a 404 for every one of them.

All 34 endpoints are implemented.

Method	Description
list_datasets	The datasets this application may read.
filter_baselines	The baselines in one dataset.
filter_baseline_categories	The categories the baselines are filed under.
get_baseline_filters	Which filters this dataset accepts for baselines.
get_baseline_filter_values	The values one baseline filter accepts.
get_baseline	One baseline in full.
filter_baseline_impacts	The impact figures for one baseline.
filter_generic_estimates	The generic estimates in one dataset.
filter_generic_estimate_categories	The categories the generic estimates are filed under.
get_generic_estimate_filters	Which filters this dataset accepts for generic estimates.
get_generic_estimate_filter_values	The values one generic estimate filter accepts.
get_generic_estimate	One generic estimate in full.
filter_generic_estimate_impacts	The impact figures for one generic estimate.
filter_product_epds	The product EPDs in one dataset.
filter_product_epd_categories	The categories the product EPDs are filed under.
filter_product_epd_statistics	The impact figures summarised across the matching product EPDs.
get_product_epd_filters	Which filters this dataset accepts for product EPDs.
get_product_epd_filter_values	The values one product EPD filter accepts.
get_product_epd	One product EPD in full.
filter_product_epd_impacts	The impact figures for one product EPD.
filter_industry_epds	The industry EPDs in one dataset.
filter_industry_epd_categories	The categories the industry EPDs are filed under.
filter_industry_epd_statistics	The impact figures summarised across the matching industry EPDs.
get_industry_epd_filters	Which filters this dataset accepts for industry EPDs.
get_industry_epd_filter_values	The values one industry EPD filter accepts.
get_industry_epd	One industry EPD in full.
filter_industry_epd_impacts	The impact figures for one industry EPD.
filter_activities	The activities in one dataset.
get_activity_filters	Which filters this dataset accepts for activities.
get_activity_filter_values	The values one activity filter accepts.
get_activity	One activity in full.
filter_activity_impacts	The impact figures for one activity.
filter_activity_exchanges	The flows into and out of one activity.
get_flow	One flow definition an exchange refers to.

Token Flex Usage Data (client.token_flex)

Reads how an Autodesk Token Flex contract is being consumed: the contract and its token pools, monthly usage totals, ad hoc queries over the raw records, and bulk CSV exports.

Every call needs a 3-legged token, and the person who authorized it must be a Token Flex administrator. Anyone else reads an empty list from list_contracts and is refused everywhere else, because every other path is keyed by a contract number. So start with list_contracts.

A 2-legged token is refused outright. Autodesk can tie a client ID to one administrator so a server can call without a user, which a Customer Success Manager arranges.

A read wants data:read and a write wants data:write. submit_query is the one POST Autodesk documents under data:read.

Queries and exports both run asynchronously.

submitted = client.token_flex.submit_query(
    contract_number,
    fields=["usageCategory", "productName"],
    metrics=["tokensConsumed"],
    where="contractYear = 1",
)
result = client.token_flex.wait_for_query(contract_number, submitted.id)
print(result.result.columns, result.result.rows)

UsageField and Metric name what a query may ask for. Autodesk documents a field marked session level in UsageField as export only. Use an export too when a query would exceed the five minute cap Autodesk puts on its running time.

export = client.token_flex.create_export_request(
    contract_number,
    fields=["usageDate", "userName", "productName"],
    metrics=["tokensConsumed"],
    usage_category=["DESKTOP_PRODUCT"],
)
client.token_flex.wait_for_export(contract_number, export.request_key)
csv = client.token_flex.download_export(contract_number, export.request_key)

An export's download link is pre-signed, so anyone holding it can read the file. It expires quickly, and refresh_export_url issues another.

update_export_schedule is a PUT and replaces the whole schedule, so read it first and send back what should stay. An empty reports schedules a run that produces no file, so allow_empty guards both writes and defaults to False.

Token Flex allows five requests a minute per user, so both wait_for_* helpers poll every fifteen seconds by default.

All 21 endpoints are implemented, plus wait_for_query, wait_for_export, and download_export.

Method	Description
list_contracts	The Token Flex contracts you administer.
get_contract	One contract, with its token pools year by year.
list_enrichment_categories	The labels a contract gave its ten custom fields.
list_enrichment_values	The distinct values under one custom field.
list_field_values	The distinct product, user, machine, or server names.
get_usage_summary	Monthly consumption totals.
get_last_usage	How current the data is, per usage and charge category.
submit_query	Submit an ad hoc query over the usage data.
get_query	Where one query has got to, and its rows once done.
wait_for_query	Poll a query until it stops.
create_export_request	Ask for the usage data as a CSV file.
list_export_requests	The export requests made in one window.
get_export_request	Where one export request has got to.
wait_for_export	Poll an export request until it stops.
download_export	The CSV file one export produced.
retry_export_request	Run a failed export request again.
refresh_export_url	A fresh download link for a finished export.
mark_export_requests_read	Mark exports read, as opening them would.
delete_export_request	Delete one export request and its file.
list_export_schedules	The recurring exports on one contract.
create_export_schedule	Create a recurring export.
get_export_schedule	One recurring export.
update_export_schedule	Replace one recurring export.
delete_export_schedule	Delete one recurring export.

Business Success Plan Reporting (client.insights)

Reads how an organisation consumes Autodesk products: seat usage for single user subscriptions and token consumption for Flex. A query returns rows of JSON, and an export writes a CSV, Excel, or JSON file.

The caller must administer a team that carries Business Success Plan benefits. list_contexts names those teams, and Autodesk describes ReportingAPIAccess as the benefit that covers this API. So start with list_contexts.

Every call needs a 3-legged token. Autodesk also accepts a 2-legged token beside a personal access token, which the caller generates on the Autodesk Account security page. Pass it as pat and the SDK sends it in ADSK-PAT.

A read wants data:read and both POSTs want data:write.

Queries and exports both run asynchronously.

submitted = client.insights.submit_query(
    fields=["productName", "usageMonth"],
    metrics=["uniqueUsers", "totalTokens"],
    where="productName LIKE 'AutoCAD%'",
    order_by="productName DESC",
)
result = client.insights.wait_for_query(submitted.id)
print(result.columns, result.rows)

UsageField and Metric name what a query may ask for. Six UsageField members carry personal data and cannot appear in a where clause. An administrator who has hashed personal data cannot request them at all.

export = client.insights.create_export(
    output_format="CSV",
    reports=["USAGE"],
    start_date=date(2025, 1, 1),
    end_date=date(2025, 6, 30),
)
client.insights.wait_for_export(export.id)
payload = client.insights.download_export(export.id, report="USAGE")

USAGE and USAGE_REPORT both need a date range, and Autodesk documents a limit of a year on it. USAGE_REPORT also needs usage_reports, and report_filters narrows it further. Each requirement raises a ValueError before the request leaves.

An export's download link is pre-signed, so anyone holding it can read the file. It expires within minutes, and download_export resolves a fresh one on every call. Autodesk keeps an export for two weeks and then deletes it.

Autodesk allows ten export requests a minute per user and ten outstanding exports, so both wait_for_* helpers poll every ten seconds by default.

All 6 endpoints are implemented, plus wait_for_query, wait_for_export, and download_export.

Method	Description
list_contexts	The teams you administer, and what each may report on.
submit_query	Submit an ad hoc query over the usage data.
get_query	Where one query has got to, and its rows once done.
wait_for_query	Poll a query until it stops.
list_exports	The exports you asked for in the last two weeks.
create_export	Ask for the usage data as a downloadable file.
get_export	Where one export has got to, and its download links.
wait_for_export	Poll an export until it stops.
download_export	One file a finished export produced.

Forma Site Design (client.forma)

Reads and writes the scene graph behind a Forma site: elements, the blobs their geometry lives in, proposals, terrain, and sun analyses.

Every path is alpha. Autodesk says an alpha endpoint can change or disappear without notice, so pin a version of this SDK before relying on it.

Forma publishes nine services behind one prefix, and each answers on its own root, so there is no single base path. The SDK reaches the eight that are not deprecated, over nine roots: Integrate answers on v1alpha and on v2alpha.

Every call but get_site names an authcontext, which is a Forma hub ID or project ID. Omitting it is a 401 even with a valid token. IDs are region specific: Forma runs in US and EMEA, so pass region for a European hub.

Autodesk designs these endpoints for an extension embedded in the Forma web client. The calling application must be registered as a service account on that extension. 15 endpoints want a 3-legged token and the other 5 accept either kind. The 8 reads want data:read and the 12 writes want data:read and data:write together.

An element is immutable. Editing one writes a new revision under a new URN, and a parent only sees the change once its own reference is rewritten.

found = client.forma.get_element(urn, "pro_abcd", recursive=True)
for child_urn, element in found.elements.items():
    print(child_urn, element.properties.get("name"))

A representation whose type is linked names a blob rather than carrying the data, so fetch it separately. get_blobs reads a batch and answers a partial result when the batch grows too large, marking the rest skipped.

mesh = found.elements[urn].representations["volumeMesh"]
glb = client.forma.get_blob(mesh.blob_id, "pro_abcd")

Writing geometry is a three step flow: reserve a link, upload the payload, then reference the blob ID it gave back.

link = client.forma.get_upload_link("pro_abcd")
client.forma.upload_payload(link.url, glb)
created = client.forma.create_element(
    "pro_abcd",
    properties={"name": "Block A", "category": "building"},
    representations={
        "volumeMesh": RepresentationInput(type="linked", blobId=link.blob_id)
    },
)

update_element merges one level deep: each named property and each named representation replaces its counterpart, and anything not named is kept. Map a name to None to remove it. children is not merged, so whatever is passed replaces the whole list.

client.forma.update_element(
    created.urn,
    "pro_abcd",
    properties={"name": "Block B", "category": None},
)

ingest_elements writes up to 1000 creates and updates at once. Items in one batch may reference each other, so a whole hierarchy goes up in one call. It reports each item separately, so read succeeded() per result rather than assuming the batch passed.

update_proposal is a replacement, not a merge. The terrain, the base, and every child become exactly what is passed. All three are required arguments, so no default can quietly empty a proposal. Read the current revision with list_revisions first.

Proposals page by cursor, and Forma reports only the next URL, so ProposalPage.next_cursor() reads the cursor back out of it.

Terrain is reserved, uploaded, and then marked. upload_terrain gzips the GLB first, which Autodesk requires. mark_terrain_uploaded takes an authcontext that its reference page does not document, because the live service answers 401 without one.

reserved = client.forma.create_terrain("pro_abcd", [[0, 0, 0], [500, 500, 90]])
client.forma.upload_terrain(reserved.presigned_s3_url, glb)
client.forma.mark_terrain_uploaded(reserved.element_id, reserved.revision, "pro_abcd")

A sun analysis runs asynchronously, so trigger_analysis returns an ID and wait_for_analysis polls it. A finished analysis links to a msgpack file of per point sunlit hours, and GroundGridFormat documents how to read it.

The deprecated Project API and the deprecated V1 write endpoints are left out. Use get_site in place of the first, and create_element, update_element, and ingest_elements in place of the second.

All 20 live endpoints are implemented, plus wait_for_analysis, upload_payload, and upload_terrain.

Method	Description
get_element	One element revision, and optionally its subtree.
get_elements	Several element revisions in one call.
get_blob	The data behind one linked representation.
get_blobs	Several blobs in one call, as a multipart body.
create_geometries	Write polygons, extruded polygons, and lines.
create_element	Write a new element into the integrate system.
update_element	Write a new revision, merging one level deep.
ingest_elements	Write up to 1000 creates and updates at once.
get_upload_link	Reserve somewhere to put a payload.
upload_payload	PUT a payload to a pre-signed Forma URL.
create_library_item	Save an element to the library.
create_proposal	Write a new proposal onto a site.
list_proposals	The proposals of one site.
update_proposal	Replace the whole content of a proposal.
list_revisions	The revisions of one proposal.
get_site	One site, and how it is projected.
get_analysis	Where one sun analysis has got to.
trigger_analysis	Start a sun analysis over an element.
wait_for_analysis	Poll a sun analysis until it stops.
create_terrain	Reserve a terrain revision and its upload link.
upload_terrain	Gzip a terrain GLB and PUT it to that link.
mark_terrain_uploaded	Tell Forma the GLB is behind the revision.
download_terrain	The GLB of one terrain revision.

Flow Graph Engine (client.flow_graph)

Runs compute jobs in the cloud, and moves the files they read and write. Autodesk aims this API at media and entertainment pipelines rather than at the AEC surfaces the rest of this client covers. Its one executor, bifrost, runs the Bifrost graphs Maya artists build.

There are two roots: jobs answer under /flow/compute/v1 and files under /flow/storage/v1. A job never carries its own data, so the order is always upload, submit, wait, download.

Either kind of token works on every endpoint. Ten job endpoints check no scope at all, create_job wants code:all, the four storage reads want data:read, and the six storage writes want data:write. Each of those was proved against the live API, one scope per token.

queue_id defaults to @default, which is the only queue Autodesk serves. A space ID is {provider}:{spaceKey}: write to scratch:@default and read a job's own results from outputs:{jobId}. Both storage providers delete a resource after 30 days, and job details go the same way.

Upload the inputs first. upload_resource runs the whole pre-signed flow: reserve the URLs, PUT each part, then complete the upload.

stored = client.flow_graph.upload_resource(
    "scratch:@default", "plane.usd", pathlib.Path("plane.usd").read_bytes()
)

A job is a list of tasks. Each task names an executor and carries that executor's own payload, and references its inputs by the URN storage gave back.

job = client.flow_graph.create_job(
    [
        TaskSpec(
            name="execute bifrost graph",
            executor="bifrost",
            inputs=[
                InputSpec(
                    source=SourceInput(uri=stored.urn),
                    target=TargetInput(path="plane.usd"),
                )
            ],
            payload={"action": "Evaluate"},
            requirements=RequirementsInput(cpu=4, memory=30720),
        )
    ],
    name="add trees",
)
finished = client.flow_graph.wait_for_job(job.id)

Running a job consumes Flow tokens. Autodesk allows 10 job submissions a minute and 60 reads a minute, and caps a task at 16 vCPUs, 122880 MB, and 48 hours.

The job lists its outputs and logs, and storage serves them. An entry names a space and a resource rather than carrying the bytes.

for output in client.flow_graph.list_job_outputs(finished.id).results:
    data = client.flow_graph.download_resource(output.space_id, output.resource_id)

list_job_updates and list_task_execution_updates follow a queue or a job by change time. Autodesk documents both after and before and refuses them together, which no reference page says, so list_job_updates catches the pair before the request leaves.

Listings page by an opaque token that Autodesk reports only inside pagination.nextUrl, so next_token() reads it back out and pagination_token sends it.

Do not send a part's entity tag when completing an upload. Autodesk's request schema demands a quoted tag and then compares it against the unquoted hash it stores, so every value it accepts answers 400. complete_upload names each part by its number, which numbered() builds, and upload_resource does the same.

All 21 documented endpoints are implemented, plus wait_for_job, upload_part, upload_resource, and download_resource.

Method	Description
list_jobs	The jobs in a queue.
list_job_updates	The jobs that changed inside one time window.
get_job	One job, with its status and progress.
create_job	Submit one or more tasks to run.
cancel_job	Stop a job that has not finished.
delete_job	Delete a finished job and the record of its run.
wait_for_job	Poll a job until it stops.
list_tasks	The tasks a job was submitted with.
list_task_executions	The runs of a job's tasks.
list_task_execution_updates	The runs that changed after one moment.
list_job_outputs	The files a job produced.
list_job_logs	The log files a job produced.
get_space	One storage space.
get_resource	The size, checksum, and URN of one stored file.
get_download_url	A pre-signed URL to read one file from.
batch_get_download_urls	Download URLs for several files in one space.
download_resource	The contents of one stored file.
get_upload_urls	Pre-signed URLs to write the parts of one file to.
batch_get_upload_urls	Upload URLs for several files in one space.
create_upload	Reserve an upload, and optionally its URLs.
batch_create_uploads	Reserve several uploads in one call.
upload_part	PUT one part to a pre-signed URL.
complete_upload	Assemble the parts into a stored resource.
batch_complete_uploads	Assemble several uploads in one call.
upload_resource	Upload a file, running the whole flow.