AI-Native Control Plane
The Model Context Protocol (MCP) is an open standard that allows AI assistants to interact with external tools. EDDI implements a comprehensive MCP server with 84 tools spanning every aspect of the platform. See the full MCP documentation for the complete tool reference.
Tool Categories
- Conversation Tools (13): Chat with agents, manage conversations, read history, pipeline logs and the per-task audit trail, agent discovery, intent-based managed chat
- Admin Tools (29): Deploy, undeploy, create, update, and delete agents, resources, triggers, schedules, and Slack or custom channel integrations
- Setup Tools (2): Create complete agents in a single call with
setup_agent, or from an OpenAPI spec withcreate_api_agent - Group Conversation Tools (18): Multi-agent discussions in 7 styles, preset templates, member follow-ups, nested groups, and standing-team backlogs
- Human-in-the-Loop Tools (10): List and inspect pending approvals, resume or cancel paused conversations, approve group phases, and submit a human member's turn
- Memory Tools (8): Persistent user memory with visibility scoping, search, and bulk operations
- GDPR Tools (2): Cascade data erasure (Art. 17) and full data export (Art. 15/20)
- Documentation Tools (2): Read EDDI's own documentation, also exposed as MCP resources and over REST
- OpenAPI 3.1: Auto-generated OpenAPI spec. Paste any external OpenAPI spec into EDDI → get a fully deployed API-calling agent
Open Standards: Not Proprietary APIs
EDDI agents can also consume external MCP tools during conversations. Configure MCP server endpoints, and agents auto-discover and invoke tools based on conversation context. Beyond MCP, EDDI implements A2A (Agent-to-Agent Protocol) for cross-platform agent communication, OpenAPI 3.1 for native spec generation and consumption, OAuth 2.0 / OIDC via Keycloak, and SSE for real-time streaming, all open standards, zero vendor lock-in. Deployed agents are also reachable through an OpenAI-compatible /v1 API, so Open WebUI, the openai SDK, LangChain, and LiteLLM can talk to them directly.
Works with Claude Desktop, Cursor, VS Code, Windsurf, Antigravity, and any MCP-compatible client. See the MCP documentation for client configuration guides.