ari1110/doc-manager-mcp

Documentation Manager

Comprehensive documentation lifecycle management powered by an MCP (Model Context Protocol) server. Automates documentation creation, maintenance, quality assessment, and synchronization for software projects.

Features

• Automatic change detection - Track code changes and affected documentation
• Link validation - Find broken links and missing assets
• Quality assessment - Evaluate docs against 7 quality criteria
• Symbol tracking - TreeSitter-based code symbol extraction (Python, Go, TypeScript, Rust)
• Dependency mapping - Automatic code-to-docs relationships
• Platform detection - Auto-detect MkDocs, Sphinx, Hugo, Docusaurus, etc.
• Documentation migration - Restructure docs with git history preservation

Installation

Claude Code Plugin (Recommended)

The doc-management plugin provides an interactive documentation workflow with specialized agents and quick commands, powered by the doc-manager MCP server.

Install:

# Add the marketplace
/plugin marketplace add ari1110/doc-manager-mcp

# Install the plugin (automatically configures MCP server)
/plugin install doc-manager@doc-manager-suite

What you get:

• @doc-expert - Documentation expert who analyzes state and directs next steps
• @doc-writer - Content specialist who creates/updates docs
• /doc-status, /doc-sync, /doc-quality - Quick commands for common workflows

Example workflow:

You: "Set up documentation management"
Claude: [Invokes @doc-expert to detect platform and initialize]

You: "/doc-sync"
Claude: [Detects changes, updates docs via @doc-writer, validates, updates baselines]

You: "Check quality before release"
Claude: [Runs quality assessment and shows actionable findings]

See the for details.

Standalone MCP Server

For using the MCP server without the plugin.

Claude Code:

claude mcp add doc-manager --scope project -- uvx doc-manager-mcp

Claude Desktop:

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "doc-manager": {
      "command": "uvx",
      "args": ["doc-manager-mcp"]
    }
  }
}

Restart your client. Your AI assistant can then use the 8 doc-manager tools directly.

Other MCP clients:

See for local development setup and alternative installation methods.

Quick Start

With Claude Code plugin:

You: "Set up documentation management for this project"
You: "/doc-sync"

With any MCP client:

Ask your AI assistant to initialize documentation management, check sync status, or validate quality. The AI will use the appropriate tools (docmgr_init, docmgr_sync, etc.) automatically.

See for complete workflows and examples.

MCP Server Tools

The doc-manager MCP server provides 8 tools for documentation management:

• docmgr_init - Initialize doc-manager for a project (modes: existing, bootstrap)
• docmgr_detect_changes - Detect code/doc changes (read-only, never writes baselines)
• docmgr_detect_platform - Auto-detect documentation platform (MkDocs, Sphinx, Hugo, etc.)
• docmgr_validate_docs - Check for broken links, missing assets, invalid code snippets
• docmgr_assess_quality - Evaluate documentation against 7 quality criteria
• docmgr_update_baseline - Update all baselines atomically (repo, symbols, dependencies)
• docmgr_sync - Orchestrate change detection + validation + quality + baseline updates
• docmgr_migrate - Restructure/migrate documentation with git history preservation

See for complete API documentation.

Architecture

Baseline System

Doc-manager maintains 3 baseline files in .doc-manager/memory/:

1. repo-baseline.json - File checksums and metadata
2. symbol-baseline.json - TreeSitter code symbols (functions, classes, config fields)
3. dependencies.json - Code-to-docs dependency mappings

Workflow:

1. docmgr_init              → Create initial baselines
2. (make code changes)
3. docmgr_detect_changes    → Detect changes (read-only)
4. (update documentation)
5. docmgr_update_baseline   → Refresh baselines

Or use docmgr_sync mode="resync" to combine steps 3-5.

Configuration

Example .doc-manager.yml:

platform: mkdocs
use_gitignore: true
exclude:
  - "tests/**"
sources:
  - "src/**/*.py"
docs_path: docs
metadata:
  language: Python
  created: '2025-01-19T20:00:00'
  version: '2.0.0'

See for all options.

Development

Running tests

# All tests
uv run pytest

# Unit tests only
uv run pytest tests/unit/

# With coverage
uv run pytest --cov=doc_manager_mcp

Running the server locally

# Install in development mode
pip install -e .

# Run server (stdio transport)
doc-manager-mcp

# Or with uv
uvx --from . doc-manager-mcp

Linting

# Ruff (linter + formatter)
uv run ruff check .
uv run ruff format .

# Pyright (type checker)
uv run pyright

Adding new tools

1. Create implementation in doc_manager_mcp/tools/
2. Define Pydantic model in doc_manager_mcp/models.py
3. Register tool in doc_manager_mcp/server.py with @mcp.tool
4. Add tests in tests/unit/ and tests/integration/
5. Update documentation

Documentation

License

MIT License