tmxcaliber

tmxcaliber is a command-line toolkit for refining, querying, and transforming TrustOnCloud ThreatModels. It helps engineers work with large model files more efficiently by filtering JSON, exporting threats and controls, mapping control objectives to frameworks, generating DFD artifacts, and comparing model versions without writing custom scripts for each task.

Highlights

• Filter ThreatModel JSON files by IDs, severity, IAM permissions, and events.
• List threats, controls, services, and feature classes from a single model or an entire directory.
• Map control objectives to SCF-supported frameworks or your own custom framework.
• Write mapping data back into a ThreatModel JSON.
• Generate threat-focused and feature-class-focused DFD XML and PNG outputs.
• Create structured change logs in JSON or Markdown.
• Scan control descriptions with regex patterns or the built-in guardduty_findings alias.

Quickstart

tmxcaliber supports Python 3.10+.

pip install tmxcaliber
tmxcaliber -h

Or run it without installing anything, using uv:

uvx tmxcaliber -h

First commands to try:

tmxcaliber list threats path/to/threatmodel.json
tmxcaliber filter path/to/threatmodel.json --severity high --output filtered.json
tmxcaliber map path/to/threatmodel.json --scf 2025.3.1 --framework-name "NIST CSF v2.0"

Installation

Install from PyPI

The recommended way to get the CLI:

pip install tmxcaliber

Or run it without installing anything, with uv:

uvx tmxcaliber -h

Pin a specific version the usual way:

pip install "tmxcaliber==<version>"

Every merge to main publishes a new version to PyPI automatically, so the latest released code is always there. The Git tags remain the release history and carry the same sdist and wheel as attachments.

Install from source

For contributors, or to work against unreleased changes. This uses uv to build the environment from the committed uv.lock:

git clone https://github.com/trustoncloud/tmxcaliber.git
cd tmxcaliber
uv sync
uv run tmxcaliber -h

If you prefer a manual virtual environment with pip:

python -m venv .venv
# PowerShell
.venv\Scripts\Activate.ps1
# bash / zsh
source .venv/bin/activate

pip install .
tmxcaliber -h

Docker

The repository includes a Dockerfile for containerized use:

docker build -t tmxcaliber .
docker run --rm -it tmxcaliber -h

Requirement note for generate

The generate command depends on drawio. If the binary is not automatically detected, provide it explicitly with --bin.

Working against the TrustOnCloud API

If your organization subscribes to TrustOnCloud, tmxcaliber can read your entitled ThreatModels directly instead of working from files you downloaded.

Set a key

Create a key in the app under Settings, Service API keys, then:

tmxcaliber init

It asks for the key without echoing it, stores it in ~/.trustoncloud/credentials readable only by you, and then calls the API to confirm it works and tell you which tenant it belongs to. Run it again any time to see what is currently set.

If TrustOnCloud has given you a different endpoint, store it with tmxcaliber init --api-url <url>. It is kept until you pass the option again.

In CI, where there is no terminal to prompt, pipe the key in:

echo "$TOC_API_KEY" | tmxcaliber init

A one-off override needs no file at all:

export TOC_API_KEY=toc-tak1-...

TOC_API_URL overrides the endpoint if TrustOnCloud has given you a different one to use. It defaults to https://api.trustoncloud.com.

The environment wins over the file, which is also why there are no profiles: one stored key, and an environment variable for the "a different one, right now" case.

The key and the endpoint are each looked up on their own. If init stored an endpoint other than the default, a key you export in TOC_API_KEY is sent to that stored endpoint unless you also set TOC_API_URL. Commands print a note on stderr when that happens, naming the endpoint and the file it came from.

The key is a TrustOnCloud credential rather than a tmxcaliber one, which is why it is named and stored this way: the same key works with curl or any other client. There is deliberately no --api-key flag, so the secret never lands in your shell history or a CI log.

Two prerequisites that are easy to miss. Your organization must have IP filtering configured, and the address you run from has to be on the allow list: an empty list denies everything, and an IPv6 caller is refused whatever the list says. Until the API is enabled for your organization, every route answers 404, including tmxcaliber me, so "not found" can mean "not enabled" rather than "does not exist".

Use a ThreatModel without downloading it

Anywhere a command takes a file, it also takes a ThreatModel id, optionally pinned to a release:

tmxcaliber list threats aws-s3
tmxcaliber filter aws-s3 --severity high --output filtered.json
tmxcaliber generate aws-s3
tmxcaliber create-change-log aws-s3@1700000000 aws-s3@1611187200

A value that exists on disk is always treated as a path, so nothing you run today changes meaning. Fetched documents are cached under ~/.tmxcaliber/cache; set TMXCALIBER_CACHE_DIR to move it.

That last example is worth noticing: comparing two released versions of the same model needs no local copies of either.

Call the API directly

The commands mirror the API's own routes, so the published reference and the CLI use the same names:

tmxcaliber threatmodels list --provider aws
tmxcaliber threatmodels get aws-s3
tmxcaliber threatmodels threats aws-s3 --feature-class S3.FC1
tmxcaliber threatmodels dfd aws-s3
tmxcaliber subscriptions list
tmxcaliber compliance frameworks list
tmxcaliber compliance mappings list --framework nist-800-53-r5
tmxcaliber ccr packs list

Collections are read in full; paging is handled for you. When the API reports that an answer is incomplete (for example, compliance mappings for ThreatModels whose SCF release is not supported yet), the command still writes what it got, names what is missing on stderr, and exits 3; pass --allow-incomplete to accept a partial answer. For anything these commands do not cover, the API takes the same key directly:

curl -H "Authorization: Bearer $TOC_API_KEY" https://api.trustoncloud.com/v1/apis

A document pulled with your key is licensed to your organization and carries its attribution. It is not a file to redistribute.

Use the API from Python

The same commands are available as a library. Each command becomes a call by turning its spaces into dots: the positional stays positional, and each option becomes a keyword argument.

from tmxcaliber import Client

api = Client()
api.me()
api.threatmodels.list(provider="aws")
api.threatmodels.threats("aws-s3", feature_class="S3.FC1")
api.compliance.mappings.list(framework="nist-800-53-r5")

Client() finds the key and endpoint exactly as the CLI does. To use a key kept somewhere else, pass it: Client(api_key=..., api_url=...). A collection comes back as a list of every row. An incomplete answer raises IncompleteAnswer, which carries the rows that did arrive, unless you pass allow_incomplete=True. Every failure is a RemoteError subclass, such as AuthenticationError or NotFound, importable from tmxcaliber. api.get(path) and api.paginate(path) remain available for a raw path.

Command Overview

Core Workflows

Filter a ThreatModel

Use filter when you want a refined ThreatModel JSON as the output.

tmxcaliber filter path/to/threatmodel.json \
  --severity high \
  --output filtered.json

Create a filtered file and a companion file containing everything removed:

tmxcaliber filter path/to/threatmodel.json \
  --permissions s3:GetObject \
  --output filtered.json \
  --output-removed

Use --exclude to invert an ID-based filter:

tmxcaliber filter path/to/threatmodel.json \
  --ids S3.T12,S3.CO1 \
  --exclude \
  --output filtered.json

List threats and controls

List threats from a single model:

tmxcaliber list threats path/to/threatmodel.json

List only high and very-high severity threats from a directory:

tmxcaliber list threats path/to/threatmodels \
  --severity high \
  --output threats.csv

List controls from a directory:

tmxcaliber list controls path/to/threatmodels --output controls.csv

List only AWS data perimeter controls:

tmxcaliber list controls path/to/threatmodels \
  --type AWS_DATA_PERIMETER \
  --output perimeter_controls.csv

List services across a directory of models:

tmxcaliber list services path/to/threatmodels --format json

List feature classes from a single ThreatModel:

tmxcaliber list feature-classes path/to/threatmodel.json

Map to a supported or custom framework

Generate a mapping to an SCF-supported framework:

tmxcaliber map path/to/threatmodel.json \
  --scf 2025.3.1 \
  --framework-name "ISO 27001 v2013" \
  --format csv

Use the exact framework name from the selected SCF version header when targeting a built-in framework.

Generate a mapping to your own framework using the starter CSVs in template/:

tmxcaliber map path/to/threatmodel.json \
  --scf 2025.3.1 \
  --framework-name "My Framework" \
  --framework-map template/myframework.csv \
  --framework-metadata template/mymetadata.csv \
  --format json

Add mapping data back into a ThreatModel

If you want the mapping persisted directly into the ThreatModel output:

tmxcaliber add-mapping path/to/threatmodel.json \
  --scf 2025.3.1 \
  --framework-name "My Framework" \
  --framework-map template/myframework.csv \
  --framework-metadata template/mymetadata.csv \
  --output enriched-threatmodel.json

Generate DFD artifacts

generate accepts either:

• a ThreatModel JSON containing base64-encoded XML in dfd.body
• a raw XML file from the main ThreatModel DFD

Generate XML and PNG outputs from JSON:

tmxcaliber generate path/to/threatmodel.json \
  --threat-dir out/threat-xml \
  --fc-dir out/feature-class-xml \
  --out-dir out/png

Generate from XML with an explicit drawio binary:

tmxcaliber generate path/to/provider_service_DFD.xml \
  --bin "C:\Program Files\draw.io\draw.io.exe" \
  --out-dir out/png

Scan control descriptions

Search control descriptions with a regex:

tmxcaliber scan path/to/threatmodel.json --pattern UnauthorizedAccess

Use the built-in GuardDuty findings alias:

tmxcaliber scan path/to/threatmodel.json --pattern guardduty_findings

Create change logs

Compare two ThreatModel versions and output JSON:

tmxcaliber create-change-log path/to/new_tm.json path/to/old_tm.json

Generate a Markdown report:

tmxcaliber create-change-log path/to/new_tm.json path/to/old_tm.json \
  --format md \
  --output changes.md

You can also scope the comparison to specific IDs and invert the selection with --exclude.

Common Examples

# List all threats across a directory
tmxcaliber list threats path/to/threatmodels

# Keep only specific controls and objectives in a filtered JSON
tmxcaliber filter path/to/threatmodel.json --ids S3.C12,S3.CO5 --output filtered.json

# Exclude known noisy threats from a threat export
tmxcaliber list threats path/to/threatmodel.json --ids S3.T2,S3.T9 --exclude

# Export services for inventory or reporting
tmxcaliber list services path/to/threatmodels --output services.csv

# Produce a custom-framework mapping
tmxcaliber map path/to/threatmodel.json --scf 2025.3.1 --framework-name "My Framework" --framework-map template/myframework.csv

# Create a Markdown change log between releases
tmxcaliber create-change-log new.json old.json --format md --output changes.md

Help Reference

Use the built-in help to inspect the full CLI surface:

tmxcaliber -h
tmxcaliber filter -h
tmxcaliber map -h
tmxcaliber add-mapping -h
tmxcaliber scan -h
tmxcaliber generate -h
tmxcaliber list threats -h
tmxcaliber list controls -h
tmxcaliber list services -h
tmxcaliber list feature-classes -h
tmxcaliber create-change-log -h

Development

Set up a local development environment with dev dependencies (Ruff, mypy, pytest, coverage, pre-commit, and type stubs):

uv sync --extra dev
uv run pre-commit install

Run the test suite:

uv run pytest -q test

Lint, format, and type-check:

uv run ruff format --check .
uv run ruff check .
uv run mypy

Contributing

Contributions are welcome. If you want to improve the CLI, add tests, or extend the documentation, open an issue or submit a pull request with a clear description of the change and its intended user impact.