Discap
Add to Discord
For developers and agents

Discap for agents: MCP server and site files

Discap is a Discord bot that records voice calls and posts a recap and transcript to the server when the call ends. The product runs inside Discord, and a server's recaps, transcripts and audio stay there, behind Discord's own permissions. This website is not an API to that data. What it offers agents is small and read-only: an MCP server with facts about Discap and its public pages, a markdown copy of any page, and a few static discovery files.

The machine-readable description is the OpenAPI 3.1 file at /openapi.json. The base URL is https://discap.bot.

Authentication and limits

None. There is no key, no token, no signup and no account, and it is free. Nothing here changes state, so there is no separate sandbox: every example below runs against production and is safe to repeat. The MCP endpoint takes one JSON-RPC message of at most 16 KiB per request. No endpoint returns a member's name, a transcript or audio. See /auth.md.

MCP server

https://discap.bot/mcp is a stateless Streamable HTTP MCP server. Send each JSON-RPC message as its own POST with Content-Type: application/json and Accept: application/json, text/event-stream. Every answer is one JSON object; the server opens no SSE stream and issues no session id, so GET answers 405. Supported protocol versions are 2025-11-25, 2025-06-18 and 2025-03-26. A browser page on another site is refused with 403. The server card is /.well-known/mcp/server-card.json.

Tools

  • get_product_info: what Discap is, who makes it, what it costs, its slash commands and its main links. No arguments.
  • list_public_pages: the pages in the site's sitemap with title and description. 50 at a time: optional section (home, product, use-cases, blog, company, legal, docs), limit (1 to 50) and offset for the next batch.
  • search_public_pages: pages whose title, description or address match the words of query (1 to 100 characters). Optional limit (1 to 20).

The tools read only the site's own public pages and /llms.txt. They never fetch another address, never read Discord servers, and offer no install, account or payment action. There are no MCP resources or prompts.

Initialize

curl -s https://discap.bot/mcp -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"curl","version":"1"}}}'
{"jsonrpc":"2.0","id":1,"result":{"protocolVersion":"2025-11-25","capabilities":{"tools":{"listChanged":false}},"serverInfo":{"name":"discap","title":"Discap website","version":"1.0.0", ...}, ...}}
curl -s -o /dev/null -w "%{http_code}\n" https://discap.bot/mcp -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" -H "MCP-Protocol-Version: 2025-11-25" -d '{"jsonrpc":"2.0","method":"notifications/initialized"}'
202

List and call tools

curl -s https://discap.bot/mcp -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" -H "MCP-Protocol-Version: 2025-11-25" -d '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'
curl -s https://discap.bot/mcp -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" -H "MCP-Protocol-Version: 2025-11-25" -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"search_public_pages","arguments":{"query":"consent"}}}'
{"jsonrpc":"2.0","id":3,"result":{"content":[{"type":"text","text":"- Recording Consent Checklist | Discap: https://discap.bot/blog/discord-recording-consent-privacy\n- ..."}],"structuredContent":{"pages":[...]}}}

Errors

An unknown method answers JSON-RPC error -32601 and an unknown tool -32602. Wrong tool arguments answer a tool result with isError: true and a short reason. Malformed JSON answers 400 with -32700; a batch or a message that is not JSON-RPC 2.0 answers 400 with -32600; a body over 16 KiB answers 413; a wrong Content-Type answers 415, and an Accept without both types answers 406.

Markdown copies of pages

GET /api/markdown?path=/about

Any page of this site as markdown (text/markdown). The path parameter is an absolute path on discap.bot; other hosts are refused. Sending Accept: text/markdown to the page URL itself gives the same answer. A missing page answers 404 with a short markdown body that points at llms.txt and the sitemap.

curl -s "https://discap.bot/api/markdown?path=/about"
curl -s -H "Accept: text/markdown" https://discap.bot/docs/api
About

# About Discap

Discap is a Discord bot that turns voice calls into notes. ...

Static files

These are plain files served as they are, not endpoints of an application.

GET /llms.txt

Plain-text product facts for language models (text/plain): what Discap does, its pages, its slash commands and the discovery files for agents.

curl -s https://discap.bot/llms.txt

GET /.well-known/api-catalog

The API catalog (RFC 9727 linkset, application/linkset+json). It links the OpenAPI file, the MCP server card, this page and llms.txt.

curl -s https://discap.bot/.well-known/api-catalog

GET /.well-known/mcp/server-card.json

The MCP server card (application/json): the endpoint, transport, protocol version, tools, and that no authentication is required.

curl -s https://discap.bot/.well-known/mcp/server-card.json

GET /.well-known/agent-skills/index.json

The agent skills index (application/json): two SKILL.md files with the sha256 digest of each.

curl -s https://discap.bot/.well-known/agent-skills/index.json

Install flow endpoints

The website also forwards /invite, /oauth/callback and a few install-flow calls to the bot. They are steps of the Add to Discord flow for a person in a browser, not tools for agents, and they are not part of this description. To add Discap to a server, send a person to /invite; Discord asks them to pick a server where they have the Manage Server permission.

Discovery files

Questions go to [email protected].