Featured

Deploy OpenClaw in 60 seconds — 20% off logoDeploy OpenClaw in 60 seconds — 20% off

Launch OpenClaw on Hostinger in about 60 seconds and keep your agent live 24/7. Our referral link gives you 20% off, no coupon code needed.

Launch on Hostinger
Run your Hermes agent on Hostinger, fully managed logoRun your Hermes agent on Hostinger, fully managed

Launch Hermes on Hostinger in one click, fully managed, no VPS knowledge needed. Use code ZACAARON10 for 10% off.

Launch on Hostinger
Crawl and scrape any site into clean data, 10% off logoCrawl and scrape any site into clean data, 10% off

Firecrawl crawls and scrapes any site into clean markdown for your agent. Get 1,000 free credits, and new users get 10% off their first purchase.

Try Firecrawl free
6,000+ web scrapers for your AI agent, start free logo6,000+ web scrapers for your AI agent, start free

Apify gives your agent live web data: 6,000+ prebuilt scrapers and actors, MCP-ready. Sign up free with $5 in usage credits.

Try Apify free
One API to scrape, enrich, and extract the internet. logoOne API to scrape, enrich, and extract the internet.

Context.dev gives your agents a single API to scrape, enrich, and extract live web data — no proxies, no parsers, no maintenance.

Start building free
SetupClaw: done-for-you OpenClaw for founders & exec teams logoSetupClaw: done-for-you OpenClaw for founders & exec teams

White-glove OpenClaw for founders and exec teams (4–50+ employees): we install, harden, integrate your tools, and maintain it — secured from day one.

Get it set up for you
SEO data APIs for your agent, $1 free credit logoSEO data APIs for your agent, $1 free credit

DataForSEO gives your agent live access to SERP results, keyword data, backlinks, and on-page SEO data through one API. New accounts get a $1 credit, good for up to 20,000 keyword or backlink lookups.

Try DataForSEO free
Reach 48,000+ AI builders

A flat monthly placement in front of developers actively installing AI tools. No lock-in, cancel anytime.

Advertise here

Works with

Claude CodeClaude DesktopCursorVS CodeClineCodex CLIOpenClaw+ any MCP client

Install to Claude Code

This server doesn't publish a one-line install command. Follow the setup in the source repository.

Summary

A documentation MCP server that crawls websites and Git repositories, stores them as Markdown, and provides tools to search and retrieve documentation for local LLMs and AI agents.

README.md

DevDoc — Documentation MCP Service

A CLI tool that crawls documentation (websites or Git repositories), saves it as Markdown, and exposes it as an MCP (Model Context Protocol) server for local LLMs and AI agents.

---

What You Can Do

With DevDoc, you can:

  1. Add public documentation from websites or Git repositories
  2. Use a built-in knowledge base of common docs without looking up URLs yourself
  3. Keep documentation cached locally for fast offline searching
  4. Search docs from the terminal before connecting any MCP client
  5. Run an MCP server for local AI tools over stdio
  6. Run an HTTP/SSE MCP server for agents and remote tooling
  7. Start the service in Docker with a single startup script
  8. Update, inspect, and remove documentation sources at any time

---

Requirements

| Tool | Notes | |------|-------| | uv | Python package manager — installed automatically by the installers | | Git | Only needed when adding git sources | | pandoc _(optional)_ | RST → Markdown conversion for Git repos; falls back to built-in converter |

---

Installation

Linux / macOS / WSL

bash install.sh

The script:

  1. Installs uv if missing
  2. Runs uv tool install --python 3.13 . → creates the devdoc binary at ~/.local/bin/devdoc
  3. Installs Playwright (Chromium) for crawl4ai web crawling
  4. Updates ~/.bashrc / ~/.zshrc if ~/.local/bin is not yet in PATH

Windows (PowerShell)

Set-ExecutionPolicy Bypass -Scope Process
.\install.ps1

The script:

  1. Installs uv via the official Windows installer if missing
  2. Runs uv tool install --python 3.13 . → creates devdoc.exe
  3. Adds the uv tools directory to your user PATH
  4. Installs Playwright (Chromium)

Manual (any platform)

uv tool install --python 3.13 /path/to/devdoc

---

Quick Start

Local CLI

# Add from the built-in knowledge base (no URL needed):
devdoc add godot
devdoc add roblox
devdoc add react

# OR supply a custom URL:
devdoc add godot https://github.com/godotengine/godot-docs.git
devdoc add mylib https://docs.example.com/

# Start the MCP server:
devdoc start

Docker

./start-docker.sh

This builds the image, starts the container, and exposes the SSE endpoint on http://localhost:8080/sse.

Use a different port if needed:

./start-docker.sh 9090

---

All Commands

Command Summary

| Command | What it does | |---------|--------------| | devdoc kb | Browse built-in documentation sources | | devdoc add | Add and download a documentation source | | devdoc list | List all configured sources | | devdoc status | Show health and environment details | | devdoc info | Inspect one source in detail | | devdoc update | Refresh one or all sources | | devdoc remove / devdoc rm | Delete a source | | devdoc search | Search local docs from the terminal | | devdoc start | Run the MCP server | | devdoc stop | Stop the background SSE daemon | | devdoc logs | View daemon logs | | devdoc mcp-config | Print MCP client configuration | | devdoc init-mcp | Attempt to configure supported MCP clients |

devdoc kb [query] [-c category]

Browse the built-in knowledge base of curated public documentation sources.

devdoc kb                    # list all (~40 entries grouped by category)
devdoc kb godot              # search by keyword
devdoc kb -c gamedev         # filter by category
devdoc kb -c frontend
devdoc kb -c backend

Categories: gamedev, frontend, backend, language, graphics, mobile, desktop, testing, api, web

Built-in sources include:

| Key | Name | Type | |-----|------|------| | godot | Godot Engine | git | | roblox | Roblox Studio | web | | unity | Unity | web | | unreal | Unreal Engine | web | | bevy | Bevy (Rust) | web | | pygame | Pygame | web | | react | React | web | | nextjs | Next.js | web | | vue | Vue.js | git | | nuxt | Nuxt | web | | svelte | Svelte | web | | astro | Astro | git | | solidjs | SolidJS | web | | tailwind | Tailwind CSS | web | | htmx | HTMX | web | | threejs | Three.js | web | | pinia | Pinia | web | | nodejs | Node.js | web | | deno | Deno | git | | bun | Bun | web | | express | Express | web | | fastapi | FastAPI | web | | django | Django | web | | laravel | Laravel | git | | supabase | Supabase | web | | prisma | Prisma | web | | graphql | GraphQL | web | | trpc | tRPC | web | | typescript | TypeScript | web | | python | Python | web | | rust | Rust (book) | git | | go | Go | web | | kotlin | Kotlin | web | | swift | Swift | web | | elixir | Elixir | web | | mdn | MDN Web Docs | web | | flutter | Flutter | web | | tauri | Tauri | web | | electron | Electron | web | | vitest | Vitest | web | | pytest | pytest | web |

---

devdoc add <name> [url]

Download and index a documentation source. URL is optional — if omitted, name is looked up in the knowledge base and fails with suggestions if not found.

# From knowledge base (recommended):
devdoc add godot
devdoc add roblox
devdoc add deno

# Custom URL (auto-detects git vs web):
devdoc add godot   https://github.com/godotengine/godot-docs.git
devdoc add mylib   https://docs.example.com/

# Options:
devdoc add <name> [url] --max-pages 200   # limit crawl size (default: 500)
devdoc add <name> [url] --delay 1.0       # seconds between requests (default: 0.5)
devdoc add <name> [url] --type web        # force source type

Downloaded files are stored in ~/.devdoc/docs/<name>/.

---

devdoc list

Show all configured sources with full details — type, document count, disk size, dates, and URL.

┏━━━━━━━━┳━━━━━━┳━━━━━━┳━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Name   ┃ Type ┃ Docs ┃ Size  ┃ Added      ┃ Updated    ┃ URL           ┃
┡━━━━━━━━╇━━━━━━╇━━━━━━╇━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ godot  │ git  │ 1842 │ 18 MB │ 2025-03-01 │ 2025-03-01 │ github.com/…  │
└────────┴──────┴──────┴───────┴────────────┴────────────┴───────────────┘

---

devdoc status

Overall health check — binary path, config location, totals across all sources, and per-source state (including Git branch/commit for git sources).

devdoc status

---

devdoc info <name>

Detailed panel for a single source: metadata, file counts, disk size, Git branch + last commit, and top-level directory structure.

devdoc info godot

---

devdoc update [name]

Re-crawl (git pull / re-crawl website) to refresh documentation. Omit name to update all sources.

devdoc update          # update everything
devdoc update godot    # update one source

---

devdoc remove <name> / devdoc rm <name>

Delete a source from the config and remove all downloaded files.

devdoc remove godot
devdoc rm godot        # short alias
devdoc rm godot --yes  # skip confirmation prompt

---

devdoc search <query>

Search documentation locally — useful for testing without an MCP client.

devdoc search "signals and slots"
devdoc search "AnimationPlayer" -s godot   # limit to one source
devdoc search "coroutine" -n 5             # top 5 results

---

devdoc start

Start the MCP server. All configured sources are available as tools.

# stdio transport (default) — for Claude Desktop, LM Studio, etc.
devdoc start

# HTTP/SSE transport — for agents that connect via HTTP
devdoc start --transport sse --port 8080 --host 0.0.0.0

Docker startup

Build and start the Docker container with the included startup script:

./start-docker.sh

Use a different port by passing it as the first argument:

./start-docker.sh 9090

The script:

  1. Builds the local Docker image
  2. Starts the container in SSE mode
  3. Maps the selected port to the same port in the container
  4. Persists DevDoc data in the Docker volume devdoc-data

The container entrypoint uses PORT and defaults to 8080.

You can also override names with environment variables:

IMAGE_NAME=my-devdoc CONTAINER_NAME=my-devdoc DATA_VOLUME=my-devdoc-data ./start-docker.sh

Useful Docker commands:

docker ps
docker logs -f devdoc
docker stop devdoc
docker start devdoc
docker rm -f devdoc
docker volume ls

---

devdoc mcp-config

Print the JSON snippet to paste into your MCP client config (e.g. Claude Desktop claude_desktop_config.json).

devdoc mcp-config              # stdio (default)
devdoc mcp-config --transport sse --port 8080

Example output: ``json { "mcpServers": { "devdoc": { "command": "/home/user/.local/bin/devdoc", "args": ["start"] } } } ``

---

MCP Tools (available to connected LLMs/agents)

| Tool | Description | |------|-------------| | list_sources | List all configured doc sources with counts | | search_docs(query, source?) | Keyword search across all docs, returns top 10 with snippets | | get_document(path) | Full content of a file (path from search results) | | list_documents(source, path?) | Browse files in a source by sub-path |

Example agent workflow

list_sources()
→ "godot (git) — 1842 docs"

search_docs("AnimationPlayer track methods")
→ 1. AnimationPlayer  classes/class_animationplayer.md
     …adds a new track to the animation…

get_document("godot/classes/class_animationplayer.md")
→ Full class reference

---

Typical Workflows

1. Add docs and search locally

devdoc add python
devdoc search "asyncio gather"

2. Add a custom documentation site

devdoc add mylib https://docs.example.com/
devdoc search "authentication token" -s mylib

3. Run for a desktop MCP client

devdoc start

Then generate client configuration:

devdoc mcp-config

4. Run as a networked SSE service

devdoc start --transport sse --host 0.0.0.0 --port 8080

5. Run in the background

devdoc start --transport sse --daemon
devdoc logs
devdoc stop

6. Run in Docker

./start-docker.sh
docker logs -f devdoc

---

Storage Layout

~/.devdoc/
├── sources.json            # source registry
└── docs/
    ├── godot/              # git clone or crawled pages
    │   ├── .git/
    │   ├── tutorials/
    │   ├── classes/
    │   │   ├── class_node.md
    │   │   └── …
    │   └── …
    └── python/
        └── …

---

Updating devdoc itself

# From the project directory:
uv tool install --reinstall --python 3.13 .

---

Notes

  • stdio mode is intended for local MCP clients such as Claude Desktop or LM Studio.
  • sse mode is intended for tools that connect over HTTP.
  • Web crawling may require Playwright browser installation.
  • Git sources work best when git is installed locally.
  • Docker mode persists DevDoc data in a named Docker volume, so your source list and downloaded docs survive container replacement.

---

Troubleshooting

| Problem | Fix | |---------|-----| | devdoc: command not found | export PATH="$HOME/.local/bin:$PATH" then restart shell | | lxml build failed | Use --python 3.13 (pre-built wheels available) | | Web crawl returns 0 pages | Run playwright install chromium manually | | Git clone fails | Check git is installed and the URL is correct | | RST files not converted | Install pandoc for best results; built-in fallback handles common cases |

See related servers & alternatives →

Related MCP servers

Browse all →

Related guides

Hand-picked reading to help you choose and use Search servers.