Skip to content

What is MCP?

The Model Context Protocol (MCP) is an open standard for connecting AI agents to external tools and data sources. Kifas exposes its full product surface (sessions, runs, projects, organizations, async tasks) as an MCP server at https://mcp.kifas.io/.

  • One integration covers many clients. Cursor, Claude Code, Continue, and any other MCP-compatible runtime can drive Kifas without us shipping per-client adapters.
  • Tools, not endpoints. MCP tools have JSON Schemas + descriptions; agents discover what’s possible without reading our docs first.
  • Streaming + async out of the box. Long-running work (recordings, large screenshots) returns a taskId; the protocol handles the polling pattern uniformly.
Client Status Install guide
Cursor Stable /mcp/install/
Claude Code Stable /mcp/install/
Continue Stable /mcp/install/
Custom (Streamable HTTP) Stable /mcp/install/

Kifas implements MCP 2025-11-01 (Streamable HTTP transport, structured tool output, OAuth 2.1 ready). Older 2025-06-18 clients are accepted via fallback.

The full reference is auto-generated from our schema source-of-truth. Start with kifas.session.create — it’s the entry point for every browser-driving workflow.