Skip to main content
Corvic exposes its capabilities through the Model Context Protocol (MCP) — an open standard that lets AI systems and applications consume tools and data in a uniform way. With MCP you can call Corvic from your own code, notebooks, agent frameworks, or backend services.

​
Two levels of access

LevelWhat it gives youGuide
Agent MCPTalk to a deployed agent — ask questions, retrieve cited resources, and trigger sync/deployAgent MCP
Room MCPProgrammatic access to a room’s workflows and feature views, including downloading their Parquet dataRoom MCP
Use Agent MCP when you want grounded answers from a configured agent. Use Room MCP when you want the underlying data and pipeline objects in a room.

​
Getting your credentials

Every MCP request needs an endpoint URL and an access token (sent as a Bearer token):
  • Agent MCP — open your agent’s Integration & Sharing → Integration tab to copy its MCP URL and Access Token. See Agents.
  • Room MCP — use the Corvic MCP endpoint with a Bearer access token and the target Room ID.
Treat access tokens like passwords. Don’t commit them to source control — load them from environment variables or a secrets manager.

​
Connect from any platform

Agent MCP works with any MCP-compatible client, so you can call Corvic from Python, notebooks, agent frameworks like OpenAI, CrewAI, or LangChain, or your own backend services.

Agent MCP

Tools for talking to a deployed agent.

Room MCP

List workflows and download feature view data.

Connect External Services

The other direction — services into Corvic.