MCP server
Early access. The MCP server ships with API access, which is being enabled for early-access accounts — join the waitlist.
The Model Context Protocol (MCP) is an open standard for connecting LLM agents to external tools and data sources. Capvo's MCP server exposes your meetings to MCP-aware clients like Claude Desktop and Cursor, so an agent can list, search, and read your transcripts as tool calls — no glue code.
Endpoint and authentication
The server speaks JSON-RPC over HTTPS at:
https://api.capvo.app/mcpAuthentication reuses your regular API key as a Bearer token — same key, same permissions, same 100 req/min rate limit as the REST API. There is no separate MCP credential to manage.
Available tools
| Tool | What it does |
|---|---|
list_meetings | List recent meetings with their AI summaries. Filter by platform or date. |
get_meeting | Fetch one meeting by UUID, full transcript included. |
search_meetings | Natural-language semantic search across your meeting summaries. |
get_action_items | Pull every action item extracted from recent meetings, filterable by owner. |
A typical agent flow: search_meetings to find the right call, get_meeting to
read the transcript, then act — draft the follow-up email, update the CRM, file
the tickets from get_action_items.
Alternative: REST + webhooks
If your stack isn't MCP-aware yet, the same data is available through the REST API and webhooks: fetch a meeting by id, list or filter your meetings, or receive the full transcript by webhook the moment a call ends, then hand it to the model of your choice.