Agents can now use real tools. Connect an MCP server, and Orqestr discovers its capabilities and makes them available to your agents.
What's included
- MCP client — Orqestr connects to external MCP servers via Streamable HTTP, discovers tools, resources, and prompts, and stores them as capabilities in your project.
- Capability discovery — on connection, all available tools are auto-discovered with their input schemas. No manual configuration needed.
- Agent-level access control — assign connections to specific agents. Control which capabilities each agent can use with three permission levels: allow, require approval, or disabled.
- Namespaced tools — capabilities are namespaced by connection name (e.g.
github_create_issue), so multiple connections can coexist without conflicts.
Error isolation
Every MCP tool call is wrapped with a timeout and error boundary. A failing tool returns an error message to the agent instead of crashing the entire run.