django-chainsaw-mcp

An MCP server and CLI that analyses a Django project rather than describing it. Not what is in here — what will hurt: what a delete takes with it, which migration breaks the pods still running, which query returns another tenant's row, what a save() sets off three hops away.

There is no model in the loop. Every answer comes from the AST and Django's own app registry, so the same input gives the same output, and nothing leaves the machine. It is an MCP server so an assistant can ask it questions, and a CLI so CI can gate on the answers.

22 checks: 19 need the app registry (16 Django, three of those DRF as well), one needs only Python, one reads FastAPI, one reads SQLAlchemy. Why this exists →

Install

It has to run in an interpreter that can import your project. Everything here reads the app registry, which means django.setup(), your settings and your apps:

/path/to/project/.venv/bin/python -m pip install django-chainsaw-mcp

A plain uvx django-chainsaw-mcp starts and then fails every check, because uvx gives it an isolated environment with no trace of your project. To avoid installing, hand uv the dependencies instead:

uvx --with-requirements requirements.txt --from django-chainsaw-mcp django-chainsaw check

Two environment variables point it at the project:

Variable	Example
`DJANGO_CHAINSAW_PROJECT_PATH`	`/srv/app` — the directory settings are importable from
`DJANGO_CHAINSAW_SETTINGS_MODULE`	`myproject.settings`

django-chainsaw project-info proves the setup before anything else, and says which half is missing. Five minutes end to end →

mcp-name: io.github.syrian963/django-chainsaw-mcp

One command

django-chainsaw check --tenant-root myapp.Organisation

44 finding(s): 1 critical, 14 high, 29 medium

CRITICAL
--------
  [deploy-safety] RemoveField drops 'legacy_code' while code still uses it
      shop/0002_remove_product_legacy_code
      During a rolling deploy the old pods keep running against the new
      schema and will fail.
      fix: Ship a release that stops using it, deploy that everywhere,
           then ship this migration.

Every analysis, merged, worst first, one exit code.

On a pull request

This is the part that decides whether a tool like this survives. Point tenancy at a five year old project and it returns two hundred candidates; nobody reads two hundred candidates, somebody adds continue-on-error, and it runs forever with nobody looking.

django-chainsaw tenancy --since main     # only what this branch changed
django-chainsaw tenancy --baseline       # everything old, ratcheted
django-chainsaw check --sarif out.json   # annotate the diff, on the line

--since compares at the merge base, so a branch that is behind main is not blamed for other people's work. --baseline keeps existing findings in the report and stops them blocking; anything new fails the build, and fixing an old one is reported so the number only ever goes down. Findings are fingerprinted on file plus identity, never the line, so adding an import does not resurrect twenty findings nobody touched.

--sarif writes the format GitHub and GitLab annotate a pull request with, so findings land on the line instead of in a log nobody opens.

baseline.md · cli.md

As an MCP server

claude mcp add django-chainsaw --scope local \
  --env DJANGO_CHAINSAW_PROJECT_PATH=/srv/app \
  --env DJANGO_CHAINSAW_SETTINGS_MODULE=myproject.settings \
  -- /srv/app/.venv/bin/python -m django_chainsaw_mcp.server

Ask it project_info first: the smallest call that proves both the transport and the Django boot. Five prompts carry the ordering the tools do not — before_deploy, why_is_this_slow, what_breaks_if_i_delete, triage, review_this_branch.

Claude Desktop, Cursor, VS Code, Windsurf, Zed, Docker →

Or as one HTML file

django-chainsaw report --out findings.html --title myproject

The HTML report, grouped by endpoint

Grouped by endpoint is the view that matters: which pages carry this, and through what call path. No server, no network, no build step — the CSS, the script and the data are all in the file, so it works from a CI artifact or an email attachment. report.md

The checks

Tool	Answers
`project_info`	Does the target project load at all? Run this first when something is broken.
`list_models`	Every model with fields, relation kind, direction and `on_delete`.
`delete_impact`	Delete one row: what cascades, what blocks, what gets nulled. Transitive.
`find_n_plus_one`	Relation traversals in a template that each cost a query, and the fix.
`scan_templates`	The same across a directory, resolving context from views.
`migration_risk`	Migrations rated: blocks writes, rewrites the table, breaks running code.
`deploy_safety`	Is this destructive migration safe to ship yet?
`find_unscoped_queries`	Which queries read data the caller may not own? The IDOR shape.
`what_happens_on`	What does this save actually trigger? Follows the signal chain.
`missing_indexes`	Fields the code filters or sorts on that carry no index.
`datetime_audit`	Naive datetimes and field defaults that break when the clock moves.
`serializer_exposure`	What DRF serializers expose, including what the next migration will add.
`serializer_nplusone`	N+1 in DRF serializers, which is where it lives in an API project.
`explain_model`	Everything about one model, plus the risks only visible combined.
`endpoint_cost`	How many queries one request costs, before anybody sends one.
`api_contract` / `api_contract_check`	What this branch changes about the API, and who it breaks.
`escaping_side_effects`	Mail and tasks fired inside a transaction that can still roll back.
`bypassed_effects`	Bulk writes that skip everything the `save()` chain promised.
`race_conditions`	Counters read into Python, changed, and saved. Also unsafe upserts.
`money_precision`	Where a decimal amount stops being exact.
`celery_arguments`	What the worker actually receives, and whether it can even be called.
`queries_in_loops`	Queries written inside a loop, split by which of three fixes applies.
`defeated_prefetches`	Prefetches paid for and then re-queried by the accessor that reads them.
`request_impact`	Every finding grouped by the entry points that reach it, so the question becomes which endpoint to fix.
`choice_typos`	Literals a field's `choices` will never match: valid SQL, zero rows, no exception.
`multiplied_aggregates`	Counts and sums multiplied by a join across two multi-valued relations.
`dangling_references`	URL names, templates, signal senders and Celery tasks nothing will resolve.
`open_endpoints`	Sensitive fields on endpoints anybody can call.
`unused_eager_loading`	Joins and prefetches nothing in the response reads.
`check`	Run everything that applies, one severity-sorted list, one exit code.
`suggest_fixes`	Findings turned into code, grouped by how safe each one is to apply.

Tool	Answers
`project_profile`	What is this built on? Counted from the project's own imports.
`blocking_in_async`	Which synchronous call stops the event loop for every request?
`fastapi_exposure`	Endpoints that serialise more than they declare.
`sqlalchemy_nplusone`	Relationships loaded one row at a time, including during serialisation.
`amplification`	Which endpoint can a stranger use to exhaust the database?

Plus the resource django://models. check profiles the project first and runs what applies, and says "does not apply, and here is why" for the rest — silence would read exactly like a clean result. Nothing about the FastAPI support imports the project, so those checks run on a checkout with no dependencies installed at all.

36 of the 37 tools declare readOnlyHint, so a client can stop asking permission for each call; the exception is api_contract_check with update=True, which writes the snapshot and says so. Every tool, argument and output shape →

What it will not tell you

Nothing here executes the target project or reads its data, which buys safety and speed and costs certainty. Every tool states its own blind spots in its own output:

delete_impact does not run signals or custom delete() overrides.
find_n_plus_one reports candidates; it reads the template and the model graph, not the queryset in the view.
migration_risk does not know row counts, PostgreSQL version, or deploy strategy.
deploy_safety cannot see getattr(obj, name), runtime SQL, or another repository. CLEAR means nothing was found here.

A confident wrong answer is worse than an incomplete one. In this kind of tooling the failure mode is not a crash, it is a plausible sentence that sends someone in the wrong direction. limitations.md

What it does to your code

It imports the target project. django.setup() imports your settings and every app in INSTALLED_APPS, and the checks additionally import the modules that declare serializers, views and URLs — so anything those do at import time happens. Do not point this at code you would not run.

It does not run your application: no view, no task, no management command. One check reads the database, read-only — MigrationLoader reads django_migrations, and nothing is written. It writes files only when you ask: fix --write applies the mechanical class of fix only, and a baseline, a contract snapshot or --sarif write where you tell them to. Nothing leaves the machine — no network calls, no telemetry, no uploads.

SECURITY.md

Documentation

docs/ is the index.


`why.md`	why this exists, three checks worth reading about, and the bar a new one clears
`quickstart.md`	five minutes from clone to first finding
`usage.md`	installing against a real project, Docker, troubleshooting
`clients.md`	Claude Code, Cursor, VS Code, Windsurf, Zed
`cli.md`	commands, exit codes, CI
`tools.md`	every tool, argument and output shape
`tested-against.md`	eighteen public projects, what they found in this tool, and the checks that never fired
`limitations.md`	what the analysis cannot see
`baseline.md`	ratcheting, so this survives a legacy codebase
`fixes.md`	suggestions as real code, and which can be applied
`architecture.md`	how it is put together, and why the bootstrap drives the design
`performance.md`	where the time goes on a large project

Contributing


`CONTRIBUTING.md`	workflow, house style, how to run the suites
`CHANGELOG.md`	every release, and the reasoning behind the changes
`SECURITY.md`	what this does to the code you point it at
`CODE_OF_CONDUCT.md`	be straight with people and be kind about it

A proposal for a new check answers four questions, which the issue template asks directly: what the defect looks like as code, how it fails in production, what already finds it, and what it must stay silent on. Two finished features were deleted from this repository after measurement showed they could not tell a real finding from a correct one.

MIT. One process stays bound to the first project it loads, because django.setup() cannot be undone — run a second instance for a second project.