doc-manager

The doc-manager MCP server provides automated documentation lifecycle tooling for software projects. It orchestrates change detection, platform detection, documentation quality assessment, dependency mapping, and migration of docs while preserving git history. Once running, your AI assistant (via Claude or other MCP clients) can invoke eight tools to initialize, scan, validate, and synchronize your documentation assets, making it easier to keep docs in sync with code and platform conventions.

Key capabilities include initializing project documentation with docmgr_init, detecting code and doc changes with docmgr_detect_changes, auto-detecting the documentation platform (MkDocs, Sphinx, Hugo, Docusaurus, etc.) via docmgr_detect_platform, validating docs for broken links or missing assets with docmgr_validate_docs, and evaluating documentation quality against seven criteria using docmgr_assess_quality. You can update baselines atomically through docmgr_update_baseline and orchestrate end-to-end workflows with docmgr_sync. For restructuring or migrating documentation while preserving history, use docmgr_migrate. These tools expose a cohesive workflow from discovery to maintenance, enabling your assistant to guide you through preparing docs for release.

To use them, interact with your MCP client to initialize the manager for a project, ask to detect changes, validate the docs, or trigger a sync cycle. The assistant will select and run the appropriate tools automatically (for example, docmgr_init followed by docmgr_sync and docmgr_assess_quality as part of a release preparation). You can also rely on the platform detection tool to tailor downstream steps to MkDocs, Sphinx, or other platforms.

Prerequisites:

• Python 3.8+ and pip
• Development tools for Python (optional but recommended): virtualenv or uv (as preferred)
• Internet access to install dependencies

Install in development mode (preferred for local development):

# Option 1: install in editable/development mode
pip install -e .

# Ensure dependencies are installed
pip install -r requirements-dev.txt

Run the MCP server locally (examples shown in README):

# Start server using uvx (from project root)
uvx --from . doc-manager-mcp

# Or install and run with uvx directly if installed as a package
uvx --from . doc-manager-mcp

If you prefer to run directly with uvx using the module name, you can also install the package and run:

uvx --from . doc-manager-mcp

Alternatively, if you are using the Claude Code plugin workflow, you can add the server via the standalone CLI snippet shown in the README:

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

Notes and tips:

• The doc-manager server maintains three baseline files under .doc-manager/memory/: repo-baseline.json, symbol-baseline.json, and dependencies.json. These are updated when you run docmgr_update_baseline.
• The standalone MCP server is designed to work with 8 tools; ensure you have the necessary dependencies and permissions for code symbol extraction (TreeSitter-based) and platform autodetection.
• If you encounter environment-related issues, verify that uvx is correctly installed and accessible in your PATH, and that Python dependencies are resolved (use a virtual environment to avoid conflicts).
• When using the Claude Code plugin workflow, commands like /doc-sync will trigger a sequence: detect changes, update docs via the writer agent, validate, and refresh baselines. Use docmgr_migrate for complex restructures that preserve history.
• For configuration, the example .doc-manager.yml shows platform, sources, and docs_path. Refer to the Configuration Reference for full options.