Django AI Boost MCP

An MCP server that exposes a Django project's models, URLs, migrations and settings to an assistant through read-only tools.

  • 114 ★ on the repository
  • Last commit
  • MIT
MCP server

Extended access

Why this level

  • query_model reads project table contents, personal data included, and passes them to the model.
  • SSE mode opens a network port protected only by a token.
All reasons and checks

vintasoftware/django-ai-boost

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add mcp/django-ai-boost

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • query_model reads project table contents, personal data included, and passes them to the model.
  • SSE mode opens a network port protected only by a token.

This is third-party code. Review the repository files before installing.

What it does

Django AI Boost runs inside the project environment and answers the assistant with data from the application itself. Tools show Django and Python versions, installed apps, models with fields and relations, URL patterns, the database schema, migrations with applied status and manage.py commands. Other calls turn a route name into a path, fetch an object's address through get_absolute_url, run Django system checks and read recent lines of file-based logs. With `query_model` the assistant selects model objects with filters, ordering and a limit: 100 objects by default, 1000 at most. An MCP prompt for documentation search builds a request with links to the Django version the project uses. The idea comes from Laravel Boost.

Who it is for. Django developers whose assistant guesses about models and routes because it cannot see the live project.

Good fit when

  • The assistant writes an ORM query and must know real model fields and relations.
  • Before a migration you need to see which ones are not applied yet.
  • A route is called by name and a ready path is needed for a test.
  • The project fails at startup and the assistant reads fresh log lines together with the system check output.

Not a fit when

  • You need writes to the database: the tools only read.
  • The project uses a framework other than Django.
  • Logs go to the console only and without a file handler there is nothing to read.

Example request

Show the blog app models with fields and relations, then the list of unapplied migrations.

MCP

Transport
stdio, sse
Authentication
not required

Environment variables

`DJANGO_SETTINGS_MODULE` required
Path to the project settings module, for example myproject.settings.
`DJANGO_MCP_AUTH_TOKEN` secret
Token for SSE mode, required when DEBUG=False.
`DJANGO_MCP_MAX_LOG_LINES`
Maximum lines read_recent_logs returns, 100 by default.

Limitations

The server needs access to project settings: `DJANGO_SETTINGS_MODULE` is required and `PYTHONPATH` is added if imports fail. Logs are read only from LOGGING file handlers, console logs are not available. On large projects some clients, PyCharm for one, truncate long list_models output, so filter apps with the `app_labels` parameter. The default stdio mode uses no authentication and no ports. SSE mode opens a network port, and with DEBUG=False the server will not start without a token. Through query_model the assistant sees table data, personal data included.

How to disable. Remove the django-ai-boost block from `.mcp.json` or the client settings and, if needed, run `pip uninstall django-ai-boost`. Stop a running SSE server separately.

FAQ

How do I connect Django AI Boost to Claude Code?

Create `.mcp.json` in the project root with the command `uv run django-ai-boost --settings myproject.settings` and `DJANGO_SETTINGS_MODULE`, then restart the extension.

Can the server change data?

No. All tools are read-only and query_model only runs selects through the ORM manager.

How do I protect SSE mode?

Set a token in `DJANGO_MCP_AUTH_TOKEN` or with the `--auth-token` flag. With DEBUG=False and SSE a token is required, and with stdio the flag causes an error.

What if list_models output gets truncated?

Pass `app_labels` with the apps you need, for example blog and auth. The answer becomes shorter than the client limit.

Where does the documentation link in the prompt come from?

The search_django_docs prompt reads the Django version from the current project and links to that version's documentation.

Discussion

No questions or reviews yet.

Official

Laravel's official MCP server, guidelines and skills: the agent sees the DB schema, logs and package versions and searches ecosystem docs

MCP serverExtended access3.7KRepository stars

A Laravel and Lumen package that adds an MCP endpoint to your app's routes and attaches your own tools to it

MCP serverExtended access331Repository stars

Five Django skills from Vinta Software: ORM and DRF, Celery, zero-downtime PostgreSQL migrations, Classy DRF classes and code review

SkillMinimal access172Repository stars

A PostgreSQL MCP server with index tuning, EXPLAIN plans, database health checks and a read-only mode

MCP serverBroad access3.5KRepository stars

A skill that interviews you about a project and writes a working Django app with up-to-date packages, settings and CI.

PluginExtended access57Repository stars

OpenAPI-MCP turns OpenAPI document operations into MCP tools with an operation allowlist and a request preview

MCP serverBroad access69Repository stars

README in short

The server is built on FastMCP and published in the official MCP registry. The README covers many clients: Cursor, Claude Desktop, Copilot, Claude Code, Cline and Zed.

README badge

Report a problem