Skip to main content

Spec-to-code mapping: core library, MCP server, and CLI

Project description

Specmap

Map AI-generated code changes back to spec intent.

Specmap generates LLM-powered annotations that link code changes to specification documents. When an AI agent writes code, reviewers can see which spec requirements each code region implements, with inline [N] citations pointing to the exact spec sections.

Quick Start

pip install specmap
specmap serve

The web UI opens in your browser. Browse PRs, generate spec-linked annotations, run AI walkthroughs, and get code reviews -- all from one interface.

How It Works

  1. Web UI (specmap serve) -- browse PRs with spec annotations overlaid on diffs. Generate AI walkthroughs that narrate the changes step by step, run code reviews with severity ratings and suggested fixes, and chat with an AI assistant about the code.
  2. MCP server integrates with your coding agent (e.g., Claude Code) to annotate code changes automatically as you work
  3. CLI generates annotations (specmap annotate), validates them in CI (specmap validate), and manages configuration
specmap serve
    │
    ├── Browse PRs with inline spec annotations
    ├── Generate annotations (lite or full mode)
    ├── AI walkthroughs (guided narrative of changes)
    ├── AI code review (P0-P4 severity, suggested fixes)
    └── Chat (investigate code with codebase tools)

Coding Agent ──MCP──► Specmap Server ──LLM──► .specmap/{branch}.json
                          │                          │
                     reads specs                committed to git
                     reads diffs                     │
                                                     ▼
                                              specmap validate (CI)

Use Specmap on Your Projects

No need to clone this repo. Install as a tool and use it in any project:

# Install from PyPI
pip install specmap

# Or via uv
uv tool install specmap

# Verify
specmap --version

Launch the web UI to review PRs with annotations, walkthroughs, and code reviews:

specmap serve

Or add the MCP server to your coding agent (.mcp.json in your project):

{
  "mcpServers": {
    "specmap": {
      "command": "specmap-mcp",
      "env": {
        "SPECMAP_API_KEY": "sk-..."
      }
    }
  }
}

See the full documentation for configuration, CLI commands, and deployment.

Features

  • Spec annotations -- LLM-generated descriptions of code regions with [N] citations to spec documents
  • Guided walkthroughs -- AI-narrated tours of PRs, configurable by familiarity level and depth
  • AI code review -- three-phase review pipeline with P0-P4 severity ratings, suggested fixes, and per-issue chat
  • Chat agent -- ask questions about code with tools for searching annotations, grepping the codebase, and reading files
  • Diff-based optimization -- incremental annotation updates proportional to the change size
  • BYOK -- bring your own key; any LLM provider via litellm (OpenAI, Anthropic, Azure, Ollama, etc.)
  • GitHub + GitLab -- auto-detected forge provider with PAT or OAuth authentication

Develop Specmap

Clone and run locally:

git clone https://github.com/jdraines/specmap.git
cd specmap

# Install dependencies
just install         # Python (core library, MCP server, CLI, API server)
just web-install     # Node (React frontend)

# Run full dev stack (API + Vite dev server)
just dev

# Run tests
just test-all        # Unit + functional tests
just lint            # ruff + tsc

# Build wheel with bundled frontend
just build

To run the web UI locally, you'll need a forge token (PAT or OAuth). See Development for the complete setup.

Prerequisites

Tool Install
Python 3.11+ System package manager
uv curl -LsSf https://astral.sh/uv/install.sh | sh
Node.js 20+ System package manager
just cargo install just or system packages
git System package manager

Project Structure

specmap/
├── src/specmap/   Python: core library, MCP server, CLI, API server
├── web/           React frontend (Vite + Tailwind)
├── tests/         Test suite (unit + functional)
├── docs/          MkDocs documentation
└── justfile       Task runner (run `just` to see all commands)

License

See LICENSE.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

specmap-0.4.0.tar.gz (462.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

specmap-0.4.0-py3-none-any.whl (312.6 kB view details)

Uploaded Python 3

File details

Details for the file specmap-0.4.0.tar.gz.

File metadata

  • Download URL: specmap-0.4.0.tar.gz
  • Upload date:
  • Size: 462.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.12

File hashes

Hashes for specmap-0.4.0.tar.gz
Algorithm Hash digest
SHA256 5b0e7e7a3cff6cc4bf62c67a2360366d0dc3712e07d7decba39b0c7c298da0eb
MD5 acb7b292fdd2963fd1e437954d800b7c
BLAKE2b-256 8a1d81cf7e4b654e90b8e8541b603580e3eb780be8d840daacb493aa574bbdc9

See more details on using hashes here.

File details

Details for the file specmap-0.4.0-py3-none-any.whl.

File metadata

  • Download URL: specmap-0.4.0-py3-none-any.whl
  • Upload date:
  • Size: 312.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.12

File hashes

Hashes for specmap-0.4.0-py3-none-any.whl
Algorithm Hash digest
SHA256 773bab26980f9456e03e91d020007b1ba4a89526f0efa29b26d19cea6df09192
MD5 1355c1bfec5d19ba35fbb79ee7b5ef2b
BLAKE2b-256 b23a2d05c64252bc8fc92b6fe089a2b92352dd6a9cb2f14ee0bfea965f20f375

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page