opengist-mcp

opengist-mcpYour gists, in the model context

An MCP server for Opengist, the self-hosted pastebin powered by Git — read, search, create, update and delete gists on your own instance.

Get started
Tools reference
GitHub

  • Fourteen tools, one API surface

    Reading, searching, writing, forking and liking gists — including revisions, commit history and raw file access — derived from the Opengist REST API and verified against a live instance.

  • Bounded by construction

    File contents are capped per file and against an overall budget, binary files are never dumped as text, and every truncation names the call that fetches the rest.

  • Register a subset, not the lot

    OPENGIST_READ_ONLY=true registers the read tools and nothing else. OPENGIST_ALLOW_TOOLS cuts finer — essential for a curated seven, your own comma-separated list, or a whole family with list_* — and OPENGIST_DENY_TOOLS subtracts. Whatever is filtered out does not exist on the protocol rather than failing when called, and a name that matches no tool stops the server at startup instead of quietly going missing.

  • Safe by default

    Deleting and publishing ask a person first, through MCP elicitation — a dialog the model cannot answer on its behalf, falling back to a server-issued token bound to the exact target and content where the client cannot show one. Upstream text is marked untrusted, and read-only mode simply does not register the write tools.

How opengist-mcp sits between an MCP client and an Opengist instanceThe MCP client speaks stdio to opengist-mcp, which calls the Opengist REST API over HTTPS with a bearer token. Every response passes through a shaping layer that bounds it and marks it as untrusted, and destructive calls are put to a person first.MCP clientClaude, Codex, …stdioopengist-mcptools14 registeredconfirmationsingle-use tokensshapingallowlist · size budgets · untrusted markersHTTPSBearer og_…Opengist/api · your instanceevery response is shaped, bounded and marked before it reaches the model
The token never leaves this process, and nothing the API returns reaches the model unshaped.

Listing the tools, narrowing them to the essential preset, and the startup abort a mistyped tool name produces

Install ​

sh
claude mcp add opengist -s user \
  -e OPENGIST_URL=https://gist.example.com \
  -e OPENGIST_TOKEN=og_your_token \
  -- npx -y opengist-mcp

Other clients — Claude Desktop, Codex, the MCP Inspector, Docker — are covered under Connecting clients.

Running it elsewhere ​

A client that cannot spawn a local process — ChatGPT connectors, Claude on the web, Cursor, LibreChat — cannot start opengist-mcp the way Claude Code does. mcp-hub is the bridge: one container serves many stdio MCP servers over Streamable HTTP, with an OAuth 2.1 login behind a single password and long-lived tokens for the clients that cannot do OAuth. Its /hub endpoint puts every server behind six meta-tools, so one connector reaches all of them without N×tool schemas in the model's context, and it speaks both protocol revisions — a question this server asks travels through it to the person at the far end instead of ending at the gateway.

Its configuration is Claude Code's mcpServers format, so the entry you already have is the entry it takes: Through mcp-hub.

Released under the MIT License.

Copyright © 2026 Willi Thiel