--- title: Doc MCP server description: Plug this documentation into Cursor, VS Code, Claude or any other MCP client. navigation: icon: i-lucide-plug --- This documentation is also served as a [Model Context Protocol](https://modelcontextprotocol.io) server. Point an AI assistant at it and it answers from the pages you are reading right now, instead of whatever it remembers about a self-hosted pastebin it saw once during training. The endpoint is one URL, over HTTP, with no account and no token: ::mcp-install :: ## What it exposes | Tool | What it does | |---|---| | `list-pages` | Lists every documentation page with its title, description and path. | | `get-page` | Returns the full markdown of one page, given its path. | Both are read-only. The server has no access to any shhh instance, no access to your pastes, and no way to create anything — it reads this documentation and nothing else. ## Adding it by hand The buttons above cover Claude, Cursor and VS Code. Everywhere else, add the URL yourself. ::code-group ```bash [Claude Code] claude mcp add --transport http shhh-docs https://shhh-docs.thoda.dev/mcp ``` ```json [.vscode/mcp.json] { "servers": { "shhh-docs": { "type": "http", "url": "https://shhh-docs.thoda.dev/mcp" } } } ``` ```json [~/.cursor/mcp.json] { "mcpServers": { "shhh-docs": { "type": "http", "url": "https://shhh-docs.thoda.dev/mcp" } } } ``` :: **ChatGPT** has no install link of its own. Custom connectors sit behind *Developer mode*, in **Settings → Apps & Connectors → Advanced**; turn it on, then add the URL as a connector. It is web-only and needs a paid plan — OpenAI moves those menu labels around, so follow [their help page](https://help.openai.com/en/articles/12584461) if they have shifted again. Other clients — Zed, Windsurf, Continue, and anything built on an MCP SDK — take the same URL as a *streamable HTTP* server; check their own documentation for where the config file lives. ::note Restart the client after editing its config. Most of them only read MCP servers at startup. :: ::warning Claude and ChatGPT reach the server from their own infrastructure, so the URL has to be publicly reachable over HTTPS. Cursor, VS Code and Claude Code connect from your own machine instead, so they also work against an instance on your network, or on a `localhost` port while you run the docs yourself. :: ## Running it from your own instance The MCP server is part of the documentation site, not of shhh itself. If you host the docs yourself with `docker/docs.Dockerfile`, the endpoint is `/mcp` on your own origin — `https://docs.example.com/mcp` — and the install buttons on this page already point at whichever host served it to you. ::warning shhh itself has **no public API and no MCP server**. An agent cannot create or read pastes on your behalf; it can only read this documentation. Both are on the [roadmap](https://github.com/thoda-dev/shhh/blob/master/ROADMAP.md) and neither is built yet. ::