PyProcore

PyProcore is the open-source automation layer for Procore built for developers, consultants, and construction teams who want typed API access, reliable file downloads, project exports, and AI-ready workflow packages without rebuilding OAuth, pagination, retries, and Procore-specific plumbing.

PyProcore handles the parts of a Procore integration that are tedious and easy to get wrong, like OAuth, token refresh, pagination, retries, typed responses, structured logging, and attachment downloads, so you work with Python objects instead of raw JSON and API plumbing.

It is built for safe read-oriented automation: project exports, local syncs, workflow packages, AI-ready review folders, and CLI-driven diagnostics without hardcoding credentials or requiring live Procore calls in tests.

---

Phase 7 Agent Layer

PyProcore is becoming an open, model-agnostic agent layer for Procore automation. The prepared 2.2.0 source adds local-first infrastructure that lets assistants discover what PyProcore can do without giving them permission to execute Procore actions.

Phase 7 includes:

• Agent Tool Registry
• Local Agent API Server
• OpenAPI / JSON Schema Export
• Agent Run Logs + Replay
• Discovery-only MCP Adapter
• Agent Evaluation Harness

This layer is discovery/spec/eval/replay infrastructure. It does not call external AI/model APIs, does not call live Procore APIs for metadata, schema, MCP discovery, replay, or eval commands, and does not enable Procore tool execution. The MCP adapter is discovery-only.

Try the local metadata and eval commands from a checkout:

PYTHONPATH=. procore-sdk agent tools
PYTHONPATH=. procore-sdk agent manifest --json
PYTHONPATH=. procore-sdk agent openapi --pretty
PYTHONPATH=. procore-sdk agent schemas --pretty
PYTHONPATH=. procore-sdk agent mcp tools --pretty
PYTHONPATH=. procore-sdk agent evals run

---

Why PyProcore

Calling the Procore REST API directly means managing the OAuth handshake, refreshing expired tokens, following pagination headers, retrying failed requests, and parsing untyped JSON on every call.

PyProcore does that once, correctly, behind a clean interface. You call a service method and get back a typed Pydantic object. It is designed as the foundation layer for higher-level tools built on Procore data: reporting, document analysis, workflow automation, and AI-powered review.

---

Features

Core SDK

• OAuth 2.0 authorization-code flow with refresh-token handling
• requests.Session HTTP client with retries, pagination, logging, and custom exceptions
• Typed Pydantic response models that serialize back to JSON
• Companies, projects, RFIs, submittals, documents, drawings, specifications, photos, and Daily Logs
• Attachment and file downloads when Procore returns usable URLs
• Human-friendly search/resolver helpers for common resources
• CLI diagnostics, auth helpers, examples, recipes, and mocked tests with no live Procore dependency

Workflow and AI-ready exports

• CSV, JSONL, Markdown, manifest, and local folder exports
• RFI, submittal, document, and project sync helpers
• Project context packages for downstream review workflows
• Enhanced RFI and submittal packages with related context
• Local AI review exports and prompt packs that do not call AI services
• JSON workflow-plan runner, scheduled automation examples, local webhook helpers, Docker templates, and CI dry-run templates

Phase 7 Agent Layer, prepared for 2.2.0

• Agent Tool Registry
• Local Agent API Server
• OpenAPI / JSON Schema Export
• Agent Run Logs + Replay
• Discovery-only MCP Adapter
• Agent Evaluation Harness
• Discovery/spec/eval/replay only; no Procore tool execution and no external model calls

---

Architecture

Package	Responsibility
pyprocore/auth/	OAuth exchange, token persistence, token refresh
pyprocore/core/	Configuration, endpoint paths, HTTP client, logging, exceptions
pyprocore/models/	Pydantic response models
pyprocore/services/	Company, project, RFI, submittal, document, drawing, specification, photo, Daily Logs, and file services
pyprocore/parser/	Email parsing utilities for future automation
tests/	Mocked unit tests with no live Procore dependency

---

Installation

Requires Python 3.12+. The latest published stable package is currently 2.1.0; the repository source is prepared for 2.2.0.

python3 -m pip install pyprocore==2.1.0

For local development:

git clone https://github.com/vibhanshu-mishra/pyprocore.git
cd pyprocore
python3 -m venv .venv
python3 -m pip install --upgrade pip
python3 -m pip install -e .

When testing unreleased local CLI changes, run commands with PYTHONPATH=. so Python resolves the local checkout before any installed copy:

PYTHONPATH=. procore-sdk --help
PYTHONPATH=. procore-sdk doctor
PYTHONPATH=. procore-sdk auth status
PYTHONPATH=. python3 -m pyprocore.app --help

---

Quick Start

python3 -m pip install pyprocore
cp .env.example .env
procore-sdk doctor
procore-sdk auth login-url

Open the login URL, approve access, then exchange the returned authorization code:

procore-sdk auth exchange-code YOUR_AUTHORIZATION_CODE
procore-sdk companies

Do not commit .env, token stores, logs, downloads, or generated workflow outputs. PyProcore redacts secrets in SDK logs, but you should still keep local configuration and OAuth tokens private.

---

Quick Example

Function style:

from pyprocore.services import list_projects

projects = list_projects(company_id=123456)

for project in projects:
    print(project.name)

Client style:

from pyprocore import Procore

client = Procore()
projects = client.projects.list(company_id=123456)

for project in projects:
    print(project.name)

---

Configuration

Copy the example file and fill in real values:

cp .env.example .env

Required variables:

PROCORE_CLIENT_ID=your_client_id
PROCORE_CLIENT_SECRET=your_client_secret_keep_private
PROCORE_REDIRECT_URI=http://localhost:8080/callback
PROCORE_LOGIN_URL=https://login.procore.com
PROCORE_API_BASE=https://api.procore.com
PROCORE_COMPANY_ID=123456

PyProcore loads .env automatically from your current working directory and does not override environment variables that are already set.

Secrets, tokens, URLs, and company IDs are never hardcoded in source.

---

Authentication

Use the auth helper commands to inspect and repair local setup:

procore-sdk doctor
procore-sdk auth login-url
procore-sdk auth exchange-code YOUR_AUTHORIZATION_CODE
procore-sdk auth status
procore-sdk auth refresh

First, run procore-sdk auth login-url and open the printed URL. After you approve access, Procore redirects to your redirect URI with a code value. Copy that value and run procore-sdk auth exchange-code YOUR_AUTHORIZATION_CODE to save tokens locally.

After that, SDK clients read the token automatically:

from pyprocore.auth.token_manager import get_access_token

access_token = get_access_token()

Expired access tokens refresh automatically whenever a refresh token is available. You can also refresh manually with procore-sdk auth refresh.

---

Usage

Once .env is configured and the one-time OAuth exchange is complete, calls return typed objects:

from pyprocore.services import list_projects

for project in list_projects(company_id=123456):
    print(project.name)

You can also use the object-oriented client interface:

from pyprocore import Procore

client = Procore()

companies = client.companies.list()
projects = client.projects.list(company_id=123456)
rfi = client.rfis.get(project_id=352338, rfi_id=102784)
open_rfis = client.rfis.list(project_id=352338, status="open")
pending_submittals = client.submittals.list(project_id=352338, status="pending")
documents = client.documents.list(project_id=352338)
documents_recursive = client.documents.list(project_id=352338, recursive=True)
document = client.documents.get(project_id=352338, document_id=456)
drawing_areas = client.drawings.list_areas(project_id=352338)
drawings = client.drawings.list(project_id=352338, drawing_area_id=123, current=True)
drawing = client.drawings.get(project_id=352338, drawing_id=789, drawing_area_id=123)
specification_sets = client.specifications.list_sets(project_id=352338)
specification_sections = client.specifications.list_sections(project_id=352338)
specification_section = client.specifications.find_section(
    project_id=352338,
    number="03 3000",
)
photo_albums = client.photos.list_albums(project_id=352338)
photos = client.photos.list(project_id=352338, album_id=123)
photo = client.photos.get(project_id=352338, photo_id=789)

Full service surface:

from pyprocore.services import (
    download_rfi_attachments,
    download_submittal_attachments,
    download_document,
    download_drawing,
    get_document,
    get_document_folder,
    get_drawing,
    get_drawing_area,
    get_rfi,
    get_submittal,
    list_companies,
    list_document_folders,
    list_documents,
    list_drawing_areas,
    list_drawing_disciplines,
    list_drawings,
    list_photo_albums,
    list_photos,
    list_projects,
    list_rfis,
    list_specification_section_revisions,
    list_specification_sections,
    list_specification_sets,
    list_submittals,
)

companies = list_companies()
projects = list_projects(company_id=123456)

rfis = list_rfis(project_id=352338)
open_rfis = list_rfis(project_id=352338, status="open")
rfi = get_rfi(project_id=352338, rfi_id=102784)
first_attachment_url = rfi.questions[0].attachments[0].url

submittals = list_submittals(project_id=352338)
pending_submittals = list_submittals(project_id=352338, status="pending")
submittal = get_submittal(project_id=352338, submittal_id=309641)

folders = list_document_folders(project_id=352338)
documents = list_documents(project_id=352338)
folder_documents = list_documents(project_id=352338, folder_id=123)
all_documents = list_documents(project_id=352338, recursive=True)
document = get_document(project_id=352338, document_id=456)
saved_document = download_document(project_id=352338, document_id=456)

drawing_areas = list_drawing_areas(project_id=352338)
drawing_disciplines = list_drawing_disciplines(project_id=352338)
drawings = list_drawings(project_id=352338, current=True)
area_drawings = list_drawings(project_id=352338, drawing_area_id=123)
drawing = get_drawing(project_id=352338, drawing_id=789, drawing_area_id=123)
saved_drawing = download_drawing(project_id=352338, drawing_id=789, drawing_area_id=123)

specification_sets = list_specification_sets(project_id=352338)
specification_sections = list_specification_sections(project_id=352338, sort="number")
specification_revisions = list_specification_section_revisions(
    project_id=352338,
    per_page=1000,
)
photo_albums = list_photo_albums(project_id=352338)
photos = list_photos(project_id=352338, album_id=123)

Documents, drawings, specifications, photos, and Daily Logs each have Procore-specific shape and permission details. See API Coverage, CLI, and the task recipes for endpoint notes, live-verification guidance, and project/company access troubleshooting.

Procore Daily Logs are organized by log type. PyProcore supports read-only counts, headers, type-specific listing, local lookup by ID/date, and grouped date summaries:

from pyprocore import Procore, list_manpower_logs

client = Procore()
counts = client.daily_logs.counts(project_id=352338)
headers = client.daily_logs.headers(project_id=352338, log_date="2026-07-10")
manpower = list_manpower_logs(project_id=352338, log_date="2026-07-10")
summary = client.daily_logs.list_for_date(
    project_id=352338,
    log_date="2026-07-10",
    log_types=["manpower", "notes", "delay"],
)

Project context packages gather available project data into a local AI-friendly folder with JSON, JSONL, Markdown summaries, a manifest, and recorded section errors:

from pyprocore.workflows import build_project_context_package

result = build_project_context_package(
    project_id=352338,
    company_id=4286480,
    output_dir="project-context",
    include=["project", "rfis", "submittals", "daily_logs"],
    max_items=100,
)
print(result.summary_path)

RFI and submittal list calls also accept optional date filters:

recent_rfis = list_rfis(project_id=352338, updated_after="2026-07-01")
recent_submittals = client.submittals.list(
    project_id=352338,
    updated_after="2026-07-01",
)

Human-friendly resolvers are available when you do not already know Procore IDs:

from pyprocore import (
    find_company,
    find_document,
    find_document_folder,
    find_drawing,
    find_drawings_contains,
    find_specification_section,
    find_project,
    find_rfi,
    find_submittal,
)
from pyprocore.services import find_project_contains

company = find_company("Tracker")
project = find_project("Sandbox Test Project")
project_by_number = find_project(number="001")
hospital_project = find_project_contains("Hospital")

rfi = find_rfi(project_id=352338, number="15")
submittal = find_submittal(project_id=352338, number="27")
folder = find_document_folder(project_id=352338, name="Drawings")
document = find_document(project_id=352338, filename="plan.pdf")
drawing = find_drawing(project_id=352338, number="S-101")
stair_drawings = find_drawings_contains(project_id=352338, text="stair")
spec_section = find_specification_section(project_id=352338, number="03 3000")

Resolvers use case-insensitive exact matching first, then partial matching. They raise NotFoundError, DuplicateMatchError, or MultipleResultsError when a lookup cannot produce exactly one typed result.

Automation Layer

The automation layer builds a single typed package for downstream workflows. It resolves project and item identifiers, fetches full metadata, optionally downloads attachments, and returns one JSON-serializable WorkflowPackage.

from pyprocore.automation import AutomationInput, build_workflow_package

package = build_workflow_package(
    AutomationInput(
        project_name="Sandbox Test Project",
        item_type="rfi",
        item_number="15",
    )
)

print(package.title)
print(package.attachments)

Convenience builders are available when you already know the workflow type:

from pyprocore.automation import build_rfi_package, build_submittal_package

rfi_package = build_rfi_package(project_id=352338, rfi_id=102784)
submittal_package = build_submittal_package(
    project_name="Sandbox Test Project",
    number="27",
    download_attachments=False,
)

Default attachment output directories are created under downloads/, for example downloads/rfi_15/ or downloads/submittal_27/. Pass output_dir to choose a custom destination.

The object client exposes the same builders:

from pyprocore import Procore

client = Procore()
package = client.automation.build_rfi_package(project_id=352338, number="15")

Workflow Automation

Workflow helpers create local files for reporting, handoff, and AI workflows. They build on the existing typed services and do not require you to request additional pages manually.

from pyprocore.workflows import export_rfis_to_csv, sync_rfis_to_folder

csv_path = export_rfis_to_csv(
    project_id=352338,
    output_path="exports/rfis.csv",
    status="open",
)

sync_result = sync_rfis_to_folder(
    project_id=352338,
    output_dir="exports/rfi-sync",
    download_attachments=True,
)

print(csv_path)
print(sync_result.manifest_path)

Available helpers:

• build_enhanced_rfi_package()
• build_enhanced_submittal_package()
• build_project_context_package()
• export_rfis_to_csv()
• export_submittals_to_csv()
• export_rfis_to_jsonl()
• export_submittals_to_jsonl()
• sync_rfis_to_folder()
• sync_submittals_to_folder()
• sync_documents_to_folder()

The object client exposes these under client.workflows.

CSV exports are best when you want a spreadsheet-friendly tracker. JSONL exports are best when another tool should process one complete typed item per line. Folder sync creates:

• rfis/, submittals/, or documents/ item folders
• item.json metadata for each item
• optional summary.md files
• a tracker CSV when enabled
• sync_manifest.json with project, item, folder, attachment, warning, and error metadata
• optional downloaded attachments

Attachment downloads skip existing files unless overwrite=True. Tracker CSV, manifest, item JSON, and Markdown summary files are regenerated when sync runs. Use dry_run=True or --dry-run to list/fetch items and preview paths without writing files or downloading attachments.

Incremental sync skips unchanged items by comparing each item ID and updated_at value against a local state file:

from pyprocore.workflows import sync_project_to_folder

result = sync_project_to_folder(
    project_id=352338,
    output_dir="exports/project-sync",
    incremental=True,
)

print(result.synced_count)
print(result.skipped_count)

Project sync combines RFIs and submittals into one output folder and writes project_sync_manifest.json plus project_sync_summary.md.

Document sync downloads project documents and writes document_tracker.csv, document_sync_manifest.json, and document_sync_summary.md:

from pyprocore.workflows import sync_documents_to_folder

result = sync_documents_to_folder(
    project_id=352338,
    output_dir="exports/documents",
    incremental=True,
    recursive=True,
)

Document downloads use download_url or url fields when Procore includes them in the file payload. Some Procore environments may require a separate secure file access step before a direct download URL is available.

Enhanced RFI packages create a read-only, AI-friendly review folder for one RFI. They can include keyword-matched related submittals, documents, drawings, specifications, photos, and Daily Logs. Downloads are off by default; related section failures are recorded by default, and generated risk flags are only possible review flags:

from pyprocore.workflows import build_enhanced_rfi_package

result = build_enhanced_rfi_package(
    project_id=352338,
    company_id=4286480,
    rfi_number="15",
    output_dir="rfi-package",
    related_sections=["drawings", "specifications", "submittals"],
    max_related_items=10,
    download_files=False,
)

print(result.review_context_path)

Enhanced submittal packages follow the same conservative pattern for submittal review. They include ai/approval_review.md for structured human review assistance and never approve, reject, or revise anything:

from pyprocore.workflows import build_enhanced_submittal_package

result = build_enhanced_submittal_package(
    project_id=352338,
    company_id=4286480,
    submittal_number="27",
    output_dir="enhanced-submittal-package",
    related_sections=["rfis", "drawings", "specifications"],
    max_related_items=10,
    download_files=False,
)

print(result.approval_review_path)

AI review exports are local-file-only. They read an existing PyProcore package folder and write an ai-export/ folder with prompts, source indexes, chunks, and checklists for downstream review tools:

from pyprocore.workflows import build_ai_review_export, build_ai_prompt_pack

review_export = build_ai_review_export("rfi-package")
prompt_pack = build_ai_prompt_pack("enhanced-submittal-package", review_type="submittal")

print(review_export.ai_review_path)
print(prompt_pack.prompt_path)

Every typed model serializes back to JSON:

json_payload = rfi.model_dump(mode="json")
json_string = rfi.model_dump_json()

---

Downloading Attachments

Attachment URLs live at:

RFI:        questions[].attachments[].url
Submittal:  attachments[].url

Download through the service functions:

rfi_files = download_rfi_attachments(project_id=352338, rfi_id=102784)
submittal_files = download_submittal_attachments(
    project_id=352338,
    submittal_id=309641,
)

The shared file service supports safe filenames, streaming writes, retries, progress logging, batch downloads, and skip-existing behavior by default:

from pyprocore.services.files import FileDownloadService

files = FileDownloadService().download_attachments(
    attachments,
    "downloads/custom",
    fallback_prefix="attachment",
    overwrite=False,
)

---

CLI

procore-sdk companies
procore-sdk projects
procore-sdk rfis --project 352338 --status open
procore-sdk submittals --project 352338 --status pending
procore-sdk find-project Hospital
procore-sdk find-rfi --project 352338 --number 15
procore-sdk download-rfi --project 352338 --id 102784
procore-sdk project-context --project 352338 --company 4286480 --output ./project-context
procore-sdk enhanced-rfi-package --project 352338 --company 4286480 --rfi-number 15
procore-sdk enhanced-submittal-package --project 352338 --company 4286480 --submittal-number 27
procore-sdk ai-review-export --package-dir ./rfi-package
procore-sdk ai-prompt-pack --package-dir ./submittal-package --review-type submittal
procore-sdk workflow-plan list
procore-sdk workflow-plan validate ./workflow.json
procore-sdk workflow-plan run ./workflow.json --dry-run
procore-sdk webhook validate examples/webhooks/rfi_created_event.json
procore-sdk agent tools
procore-sdk auth status

Most CLI commands print formatted JSON. Export and sync workflow commands print short human-readable summaries. See CLI for the full command reference.

---

Examples and Recipes

Runnable example scripts live in examples/. They show common SDK tasks such as listing projects, fetching RFIs, downloading attachments, documents, drawings, specification revisions, photos, and Daily Logs, building workflow packages, exporting CSVs, syncing local review folders, building AI-ready project context packages, and creating enhanced RFI review and submittal review packages. The local AI export examples show how to turn an existing package folder into prompt, chunk, source-index, and checklist files without calling Procore or an AI service.

Local workflow plans let you repeat existing PyProcore workflows from JSON files for scheduled syncs, repeatable exports, and AI review package generation. Plans do not execute shell commands and do not mutate Procore data. Validate and dry-run plans before scheduling them:

procore-sdk workflow-plan list
procore-sdk workflow-plan validate examples/workflow_plans/project_context_and_ai_export.json
procore-sdk workflow-plan run examples/workflow_plans/project_context_and_ai_export.json --dry-run

Scheduled automation examples are available for cron, macOS launchd, Windows Task Scheduler, GitHub Actions, and local shell/PowerShell runners. Start with the templates in examples/scheduled/, examples/github-actions/, and the Schedule a Workflow Plan recipe.

Examples can be syntax-checked without credentials or live Procore access:

make examples-check

Task-based guides live in docs/recipes/. Recipes explain when to use each pattern, which environment variables are needed, what output to expect, and how to troubleshoot beginner-friendly issues.

For AI-ready RFI review workflows, start with Build Enhanced RFI Package or RFI AI Review Context.

For AI-ready submittal review workflows, start with Build Enhanced Submittal Package or Submittal AI Review Context.

For local-file-only AI review exports, start with Build an AI Review Export, Build an AI Prompt Pack, or Review a Procore Package with AI.

For local automation plans, start with Validate a Workflow Plan, Run a Workflow Plan, or Workflow Plan Examples.

For scheduled automation, see Schedule a Workflow Plan, Cron, macOS launchd, Windows Task Scheduler, or GitHub Actions.

Local webhook helpers can validate, redact, save, list, and dry-run dispatch sample webhook JSON payloads without running a hosted webhook server. Start with Local Webhook Helpers, Validate a Webhook Event, Save a Webhook Event, or Dispatch a Webhook Event to a Workflow Plan.

The prepared 2.2.0 source includes Phase 7 agent-layer infrastructure: Agent Tool Registry, Local Agent API Server, OpenAPI / JSON Schema Export, Agent Run Logs + Replay, Discovery-only MCP Adapter, and Agent Evaluation Harness. These commands are metadata/spec/eval/replay infrastructure only. They do not execute Procore tools, do not call external AI/model APIs, and do not require Procore credentials unless a separate live SDK workflow is run. MCP tool execution remains disabled. The older lowercase command names map to the same Phase 7 components: local agent API server, OpenAPI and JSON Schema exports, agent run logs and replay, a discovery-only MCP adapter, and a local deterministic eval harness.

procore-sdk agent tools
procore-sdk agent tool procore.find_rfi
procore-sdk agent manifest --json
procore-sdk agent openapi --pretty
procore-sdk agent schemas --pretty
procore-sdk agent serve --port 8765
procore-sdk agent serve --run-log-dir agent-runs
procore-sdk agent runs replay RUN_ID --run-log-dir agent-runs
procore-sdk agent mcp tools --pretty
procore-sdk agent mcp manifest --pretty
procore-sdk agent evals list
procore-sdk agent evals run

See Agent API and Use the Agent Tool Registry, or Run the Local Agent API Server, or Export Agent OpenAPI, Inspect Agent Run Logs, or Replay Agent Run, Export Agent MCP Tools, or Run MCP Discovery Adapter, or Run Agent Evals.

Docker and CI templates are optional. They help teams dry-run workflow plans in repeatable local or GitHub Actions environments without committing secrets or generated outputs. Start with Docker Automation, CI Automation, or examples/docker/.

Release guidance lives in docs/release.md. It covers versioning, local package validation, changelog updates, and manual publishing checklists. The current public release status is captured in docs/project-status.md.

---

Documentation Site

PyProcore now includes an MkDocs-ready documentation site structure under docs/. The docs are organized for local browsing and future publishing, but they are not published as a hosted site yet.

Install the optional docs tooling when you want to preview or build the site:

python3 -m pip install -e ".[docs]"
make docs-serve
make docs-build

The MkDocs site links the getting started guide, authentication help, CLI usage, API coverage, workflows, AI review guidance, automation docs, recipes, contributing guidance, release guide, and changelog.

Detailed endpoint notes, live smoke helper commands, and 403 troubleshooting live in API Coverage and CLI. The README stays focused on installation, safe usage, and where to go next.

---

Pagination

Collection methods use ProcoreClient.get_all(), which follows Procore pagination headers automatically. Business logic should call the service method or get_all() directly and never request page 2 by hand.

---

Logging

Structured logs are written to:

logs/sdk.log
logs/errors.log

Request logs record method, endpoint, status, elapsed time, and retry count. Exception logs record stack traces, exception type, request URL, HTTP status, and response body when available. The logger redacts sensitive keys such as authorization headers, access tokens, refresh tokens, and client secrets.

---

Security Checks

PyProcore includes local checks to help avoid committing credentials or generated project data. Run the secret scanner before opening a pull request:

make secret-check

For a broader local quality gate, run:

make quality-check

Optional pre-commit hooks are available for whitespace, file format, private-key, Black, and isort checks:

python3 -m pip install pre-commit
pre-commit install
pre-commit run --all-files

Never commit .env files, token stores, Authorization headers, access tokens, refresh tokens, client secrets, logs containing credentials, downloads, or private project data. See docs/security.md and SECURITY.md for details.

---

API Coverage

See API Coverage for implemented resource families, endpoint notes, live-verification limitations, and Procore permission context.

---

Roadmap

Released in 2.1.0

• Core SDK: OAuth, token refresh, retries, pagination, logging, typed models, and custom exceptions
• Procore read coverage: companies, projects, RFIs, submittals, documents, drawings, specifications, photos, and Daily Logs
• Human-friendly resolvers and object-oriented client helpers
• Attachment/file downloads where Procore returns usable URLs
• Workflow exports: CSV, JSONL, sync folders, project context packages, enhanced RFI packages, enhanced submittal packages, and local AI review exports
• Automation foundation: workflow plans, scheduled examples, webhook helpers, Docker templates, CI dry-run examples, documentation site, recipes, security docs, and release tooling

Phase 7 Agent Layer

PyProcore is becoming an open, model-agnostic agent layer for Procore automation. The prepared 2.2.0 source adds local-first infrastructure that lets assistants discover what PyProcore can do without giving them permission to execute Procore actions.

• Agent Tool Registry
• Local Agent API Server
• OpenAPI / JSON Schema Export
• Agent Run Logs + Replay
• Discovery-only MCP Adapter
• Agent Evaluation Harness

This layer is discovery/spec/eval/replay infrastructure. It does not call external AI/model APIs, does not call live Procore APIs for metadata, schema, MCP discovery, replay, or eval commands, and does not enable Procore tool execution. The MCP adapter is discovery-only.

Try the local metadata and eval commands from a checkout:

PYTHONPATH=. procore-sdk agent tools
PYTHONPATH=. procore-sdk agent manifest --json
PYTHONPATH=. procore-sdk agent openapi --pretty
PYTHONPATH=. procore-sdk agent schemas --pretty
PYTHONPATH=. procore-sdk agent mcp tools --pretty
PYTHONPATH=. procore-sdk agent evals run

Future

Phase 8 — Guarded Tool Execution

• guarded execution design
• human approval gates
• write-action safety model
• real MCP execution after safety design

Phase 9 — Expanded Procore + Platform Layer

• async client
• Observations
• Correspondence
• plugin architecture

Phase 10 — AI Workflows, Evals, and Deployment

• vector DB examples
• engineering assistant examples
• golden datasets and model evals
• private deployment patterns
• richer MCP integration

---

Diagnostics

Run:

procore-sdk doctor

This checks local configuration, token storage, Python version, and writable SDK folders without making live Procore calls.

For an authenticated Procore connectivity check:

procore-sdk doctor --live

---

Troubleshooting

Error	Likely cause and fix
ConfigurationError	.env is missing or a required key is absent.
AuthenticationError	Complete the first OAuth code exchange; confirm pyprocore/auth/token_store.json holds a refresh token.
AuthorizationError	The Procore user lacks access to the target company, project, or resource.
ResourceNotFoundError	Project, RFI, or submittal ID is wrong for the configured company.
Attachments not downloading	Check logs/errors.log for HTTP status and response body. Existing files are skipped unless overwrite=True.

---

Tests

Run unit tests:

make test

Run coverage:

make coverage

---

Development

Clone the repository and install the package in editable mode:

git clone https://github.com/vibhanshu-mishra/pyprocore.git
cd pyprocore

python3 -m venv .venv
source .venv/bin/activate

pip install -e ".[dev]"

If you're working in a restricted environment where build isolation cannot access PyPI, install with:

pip install --no-build-isolation -e ".[dev]"

When testing unreleased local CLI changes, run commands with PYTHONPATH=. so Python resolves the local checkout before any installed copy:

PYTHONPATH=. procore-sdk --help
PYTHONPATH=. procore-sdk doctor
PYTHONPATH=. procore-sdk auth status
PYTHONPATH=. python3 -m pyprocore.app --help

Useful development commands:

make lint
make typecheck
make test
make coverage

---

Project Resources

• Changelog
• Contributing
• Security
• Support
• Code of Conduct
• Architecture notes
• Roadmap
• Suggested GitHub labels
• Documentation site
• Security guide
• Project status
• Examples
• Recipes

---

Supported Environments

• Procore Production
• Procore Sandbox

---

👷 About

Built by Vibhanshu Mishra, PE — Structural Engineer at AG&E Structural Engenuity, Austin, TX.

Specialising in steel and mission-critical structures, with a side focus on miscellaneous structures — stairs, platforms, equipment supports. Building AI and automation tools for a niche that nobody else is covering yet.

• 🔗 RISA-3D MCP Server — Connect Claude AI to your RISA-3D structural models
• 🔗 TSD MCP Server — Connect Claude AI to your TSD structural models

---

Contributing and Support

Contributions, issues, and feature requests are welcome. Start with CONTRIBUTING.md, SUPPORT.md, SECURITY.md, and CODE_OF_CONDUCT.md.

Please open an issue before submitting large changes, and never paste Procore client secrets, access tokens, refresh tokens, .env values, token stores, Authorization headers, or private project data into public issues or pull requests.

---

License

Released under the MIT License. See LICENSE for details.

---

Disclaimer : PyProcore is an independent open-source project and is not affiliated with or endorsed by Procore Technologies.