Agents
AI agents combine LLM inference with MCP tool access in a single request. Agents are executed via the inference endpoints using your platform or end-user API key.
Agent management endpoints (create, list, update, delete agents) are not yet available via the API. Agent execution runs through Chat Completions with tools enabled.
How Agents Work Today
1.
Platform admin activates MCP apps via POST api.assistiv.ai/v1/platforms/{id}/mcp/configs.
2.
End user connects apps via the OAuth flow at mcp.assistiv.ai/oauth/authorize.
3.
Use POST mcp.assistiv.ai/mcp with JSON-RPC 2.0 to execute tools directly, or pass tool definitions into POST api.assistiv.ai/v1/chat/completions for an LLM-driven agent loop.