Citadel MCP
A Code Mode MCP server that provides AI coding agents with curated, per-stack documentation as typed tools, enabling efficient document access and search across multiple stacks.
README
Citadel MCP
A Code Mode MCP server that gives AI coding agents curated, per-stack documentation as typed tools.
Citadel is built to be forked and tailored. The hosted package ships with a small starter set of stacks, but the real value comes from packaging the docs your agents actually need. Each stack is a tiny module — adding one is mostly mechanical.
Getting Started
Requirements
- Node.js v20.19 or a newer latest maintenance LTS version
- npm or pnpm
Install with add-mcp
Install the MCP server for all your coding agents:
npx add-mcp citadel-mcp@latest
Add -y to skip the confirmation prompt and install to all detected agents already in use in the project directory. Add -g to install globally across all projects.
Manual installation
Add the following config to your MCP client:
{
"mcpServers": {
"citadel": {
"command": "npx",
"args": ["-y", "citadel-mcp@latest"]
}
}
}
[!NOTE] Using
citadel-mcp@latestensures that your MCP client will always use the latest version of the Citadel MCP server.
MCP Client Configuration
<details> <summary>Amp</summary>
Using Amp CLI:
amp mcp add citadel -- npx citadel-mcp@latest
Or configure manually:
Follow Amp's MCP documentation and apply the standard configuration shown above.
</details>
<details> <summary>Claude Code</summary>
Use the Claude Code CLI to add the Citadel MCP server:
claude mcp add -s user citadel -- npx -y citadel-mcp@latest
Use -s project instead of -s user to scope the install to the current project. Restart Claude Code, then verify with claude mcp list. A working server advertises a single tool named docs.
</details>
<details> <summary>Codex</summary>
Using Codex CLI:
codex mcp add citadel -- npx citadel-mcp@latest
Or configure manually:
Follow the MCP setup guide with the standard configuration format:
- Command:
npx - Arguments:
-y, citadel-mcp@latest
</details>
<details> <summary>Cursor</summary>
Go to Cursor Settings -> MCP -> New MCP Server. Use the JSON config provided above.
</details>
<details> <summary>Gemini</summary>
Using Gemini CLI:
Project-wide installation:
gemini mcp add citadel npx citadel-mcp@latest
Global installation:
gemini mcp add -s user citadel npx citadel-mcp@latest
</details>
<details> <summary>VS Code / Copilot</summary>
Using VS Code CLI:
code --add-mcp '{"name":"citadel","command":"npx","args":["-y","citadel-mcp@latest"]}'
Or configure manually:
Follow the official VS Code MCP server setup guide and add the Citadel server through VS Code settings.
</details>
<details> <summary>Warp</summary>
Navigate to Settings | AI | Manage MCP Servers and select + Add to register a new MCP server with the following configuration:
- Name:
citadel - Command:
npx - Arguments:
-y, citadel-mcp@latest
</details>
What you get out of the box
The hosted package includes a starter set of stacks so you can try it immediately:
- Next.js (App Router, Next.js 16)
- React
- Turborepo
- Supabase guides
This list is intentionally small. For the best results, fork this repo and add the stacks your agents care about — your internal libraries, the framework version you actually use, or any docs site that publishes machine-readable markdown. See docs/adding-a-docs-tool.md.
How it works
Citadel is a Code Mode server: instead of advertising one tool per docs source, it advertises a single docs tool. The agent writes one async () => { ... } per turn that calls codemode.<stack>_docs(...) and codemode.<stack>_index() directly, and the server runs that code in a local Node sandbox. N doc fetches collapse into one round-trip.
Each stack contributes two callable tools to the sandbox SDK:
<stack>_index()— returns the doc index so the agent can pick a valid path<stack>_docs({ path })— fetches that specific doc as markdown
Citadel also exposes a cross-stack ranked search: codemode.docs_search({ query, stacks?, fetch: true }) returns BM25-ranked matches with markdown content attached in a single call.
Documentation
- Using Citadel as an AI agent — playbook for the
docstool, served at runtime as MCP resourcecitadel://docs/agent-usage. SDK reference, parallel fan-out, error handling, worked examples. - Adding a docs tool — package a new stack as
<stack>_index+<stack>_docs.
Local Development
To run the MCP server locally for development:
-
Clone the repository.
-
Install and build:
pnpm install pnpm build -
Configure your MCP client to use the local version:
{ "mcpServers": { "citadel-dev": { "command": "node", "args": ["/absolute/path/to/citadel-mcp/dist/index.js"] } } }Or with the Claude Code CLI:
claude mcp add -s user citadel-dev -- node /absolute/path/to/citadel-mcp/dist/index.js
For development with auto-reload, point the MCP client at tsx and the src entry instead:
claude mcp add -s user citadel-dev -- npx tsx /absolute/path/to/citadel-mcp/src/index.ts
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.