django-orm-lens

django-orm-lens

Static analysis + MCP server for Django models. Terminal- and AI-agent-friendly.

Listed in the official MCP Registry as io.github.FROWNINGdev/django-orm-lens.

Ships with a zero-dependency parser, a JSON/Markdown/table CLI, and an optional MCP (Model Context Protocol) server so any AI coding agent — Cursor, Aider, Continue, and any other MCP client — can navigate your Django schema without importing Django or spinning up your app.

Companion to the Django ORM Lens VS Code extension.

Paid-tier capabilities, free and MIT

Schema review is a paid category nearly everywhere. A bot that reviews every pull request, analysis that follows a queryset past the function it was built in, a check that catches schema drift, index advice grounded in real table statistics — those normally sit behind a per-seat or per-database subscription.

All of it is here, MIT-licensed, with no tier gate, no seat count, no account and no telemetry:

Capability usually sold as a paid tier	Here
PR review bot for schema changes — posts once, then updates in place	blast-radius + the GitHub Action
Analysis that follows a queryset across functions	nplusone
Schema drift detection	drift
Index proposals from observed QuerySet usage	suggest-indexes
Migration risk weighed against real table sizes	blast-radius --stats
Blast radius of a destructive migration	blast-radius
Cross-layer impact of removing a field	impact

There is no Pro tier, and none is planned.

Install

# Core CLI (zero third-party deps)
pip install django-orm-lens

# With the MCP server (adds the `mcp` package)
pip install "django-orm-lens[mcp]"

Requires Python 3.9+. Works on Linux, macOS, and Windows.

CLI usage

# Scan a Django project for models (JSON, Markdown, or table)
django-orm-lens scan -f json
django-orm-lens scan -f markdown
django-orm-lens scan -f table

# Describe one model
django-orm-lens describe blog.Post
django-orm-lens describe Post -f json

# Compact hover card (great for pipeing into your editor)
django-orm-lens hover blog.Post

# Flat list — pipes into fzf, grep, etc.
django-orm-lens list | fzf

# ER diagram — Mermaid (default), DBML, D2, or PlantUML
django-orm-lens er > schema.mmd
django-orm-lens er -f dbml > schema.dbml     # paste into dbdiagram.io
django-orm-lens er -f d2 > schema.d2         # render: d2 schema.d2 schema.svg
django-orm-lens er -f plantuml > schema.puml

# Diff two schema dumps (exit 1 on changes — CI-friendly)
django-orm-lens diff before.json after.json

# Static analyzers — text, json, sarif, or github annotation output
django-orm-lens nplusone --format github     # N+1 findings as PR annotations
django-orm-lens migration-risk -f sarif      # SARIF for GitHub Code Scanning
django-orm-lens suggest-indexes blog.Post    # Meta.indexes proposals from usage
django-orm-lens signals                      # sender→signal→handler graph
django-orm-lens migration-deps blog -f mermaid   # per-app migration DAG
django-orm-lens cascade blog.Author          # delete blast radius by on_delete

Every command accepts --path <dir> and repeatable --exclude <glob>. Defaults skip migrations/, venv/, .venv/, env/, and node_modules/.

MCP server — for AI coding agents

The MCP server exposes ten read-only tools that any MCP-compatible agent can call while it edits your Django project:

Tool	Purpose
list_apps	Every Django app in the workspace with model counts
list_models	Flat app.Model list, optional app filter
describe_model	Full field / relation / Meta detail for one model
find_relations	Inbound + outbound relations for one model
cascade_preview	Blast radius of one delete(), grouped by on_delete
er_diagram	ER diagram — mermaid / dbml / d2 / plantuml
describe_migration_dependency	Per-app migration DAG: roots, leaves, cross-app deps
suggest_indexes	Meta.indexes proposals from observed QuerySet usage
signal_graph	Sender→signal→handler graph from @receiver decorators
nplusone_scan	Static N+1 findings for the whole workspace

Start the server manually

django-orm-lens-mcp     # dedicated entry point
# or
django-orm-lens mcp     # subcommand

Workspace resolution (py-1.3.0+). Priority: explicit workspace_root argument on the tool call → DJANGO_ORM_LENS_ROOT env var → current working directory. If none resolves to a Django project (manage.py, django in pyproject.toml, or any models.py) you get a structured error envelope back — {"error": "WORKSPACE_NOT_DJANGO", "hint": "…"} — instead of an empty list, so the agent knows what to do next.

Optional sandbox: set DJANGO_ORM_LENS_ALLOWED_ROOTS (;-separated on Windows, :-separated elsewhere) to a whitelist of prefixes; any path outside them is rejected with WORKSPACE_NOT_ALLOWED.

Register it with an agent

Cursor — add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "django-orm-lens": {
      "command": "django-orm-lens-mcp",
      "env": { "DJANGO_ORM_LENS_ROOT": "/abs/path/to/your/project" }
    }
  }
}

Any MCP client — same shape, generic tool. Point command at the installed django-orm-lens-mcp binary. Two ways to tell it which Django project to scan:

• Set DJANGO_ORM_LENS_ROOT in env — the whole session uses one project. Simplest for single-repo workflows.
• Pass workspace_root on each tool call — the agent switches projects per call. Useful for mono-repos and multi-workspace setups.

The tool signatures include workspace_root: str = "" as an optional parameter, so any agent inspecting tools/list sees it and can supply it.

CI gates

diff and nplusone exit 1 on findings; migration-risk exits 1 on critical findings (--exit-zero for report-only). Two annotation-ready formats: --format github prints ::warning/::error workflow commands (PR annotations with no extra permissions), --format sarif emits SARIF 2.1.0 for github/codeql-action/upload-sarif.

pre-commit users get two ready-made hooks:

repos:
  - repo: https://github.com/FROWNINGdev/django-orm-lens
    rev: py-v1.8.1
    hooks:
      - id: django-orm-lens-nplusone
      - id: django-orm-lens-migration-risk

GitHub Actions users get a composite action: uses: FROWNINGdev/django-orm-lens@action-v1 with command: / format: inputs.

Why?

Django's ORM is Python, and Python is dynamic. AI agents that only see models.py as raw text miss:

• which fields belong to which model;
• the direction and cardinality of every relation;
• what Meta.ordering, unique_together, and constraints actually contain;
• which app owns which model when the project uses split models/ packages.

django-orm-lens gives them a static, deterministic, JSON view of the schema — no Django boot, no database, no side effects. And you get a nice CLI for humans too.

Programmatic API

from django_orm_lens import scan_workspace

index = scan_workspace(".")
for app in index.apps:
    for model in app.models:
        print(f"{app.name}.{model.name} — {len(model.fields)} fields")

# The full parsed tree serialises to the same JSON schema the VS Code
# extension emits, so tools can share it interchangeably.
import json
json.dumps(index.to_dict(), indent=2)

Links

• Repo: https://github.com/FROWNINGdev/django-orm-lens
• Issues: https://github.com/FROWNINGdev/django-orm-lens/issues
• Changelog: https://github.com/FROWNINGdev/django-orm-lens/blob/main/CHANGELOG.md
• VS Code extension: https://marketplace.visualstudio.com/items?itemName=frowningdev.django-orm-lens

MIT licensed.