Servidor MCP de Wajo

• Direcciónhttps://mcp.wajo.ai/mcpno está en el registro oficial: respondió al preguntarle

Ninguna entrada del registro oficial de MCP nombra este servidor. Respondió en wajo.ai cuando el Atlas preguntó en la dirección donde suelen montarse los servidores, así que es del propio proveedor. Los clientes y catálogos que listan servidores a partir del registro no lo muestran.

• Estadorespondecomprobado el 2026-10-27
• Herramientas21
• El servidor se llamawajo-agentmcp
• Versión del servidor0.1.0
• Versión del protocolo2025-06-18

El 2026-10-27 el servidor respondió a initialize y tools/list sin iniciar sesión. Primera respuesta en 73 ms.

• Aún no se ha juzgado si sus herramientas hacen el trabajo del producto.

Lo dicen reglas a partir de las mediciones de arriba. La primera llamada probable la elige un modelo de la lista de herramientas; lee su descripción antes de fiarte.

• Nota INITE MCP78 de 100responde 100 · definiciones 100 · contexto 83 · conexión 100 · rebajada por las señales de aviso de abajo
• Respondióel 100% de los días7 de 7 comprobaciones diarias
• Mediana de la primera respuesta116 ms
• Espacio en el contexto del agente≈ 5671 tokenslas definiciones de todas sus herramientas (21), que el agente lleva en cada turno
• Definiciones de herramientasA · 100 de 100
• Resultados tipados0 de 21 herramientasdeclaran outputSchema: el resultado llega tipado y se lee de una vez

• Todas las herramientas pasan las seis comprobaciones.

Un programa comprueba el servidor una vez al día: initialize y tools/list, sin modelo. Los tokens son la longitud de las definiciones, a unos cuatro caracteres por token. La nota promedia seis comprobaciones del nombre, la descripción, el esquema y las anotaciones de cada herramienta; dice lo bien descritas que están, no si funcionan. La nota pondera solo lo medido: si responde (25%), si hace el trabajo del producto (25%), definiciones (20%), contexto (10%), uso (10%) y lo fácil que es conectarse (10%). Lo que no se midió queda fuera, y una instrucción oculta limita la nota a 40. El estado de los servidores MCP de los proveedores: método y números →

• build_agent

  Chat with the built-in Agent Builder to design and persist a new AgentSpec. Mint session_id via generate_session_id (with agent_id = "b4409947-a46d-497e-94bb-fbba525185b7") first; pass the same session_id across turns to continue the…

• build_connector

  Register an HTTP API as a connector the user's agents call through connector_call. Use it when the user needs a service that has an HTTP API but no existing integration; when the vendor publishes an official MCP server, use…

• build_mcp_server

  Register a remote MCP server so the user's agents can call its tools via the `mcp_call` runtime tool. Takes an MCP server URL, optional auth metadata, and persists an MCPServerSpec with the discovered tools[] catalogue cached. The call…

• chat_agent

  Chat with the persisted AgentSpec bound to session_id. Mint session_id via generate_session_id first; chat_agent does not accept agent_id. Pass the same session_id across turns to continue the conversation. Long turns may return…

• copy_agent

  Clone an existing agent into another project — including one in a different workspace/org — producing a fresh, independently-editable copy you own. The source agent is left untouched; the copy gets a new id, the destination project, you…

• delete_agent

  Remove the AgentSpec at id. Returns ok=true on success; a missing or inaccessible id returns an error result with "agent X not found". Deleting requires the agent's owner or an org admin. Confirm the id with the user before calling.

• delete_connector

  Delete the ConnectorSpec at connector_id. Cross-tenant delete returns "connector X not found". Confirm the connector_id with the user before calling — this is destructive.

• delete_session

  Delete the MCP session at session_id, including the binding and any underlying conversation state. Cross-tenant delete returns "session X not found". Confirm the session_id with the user before calling — this is destructive.

• delete_webhook

  Delete one of your webhooks by id (from list_webhooks or register_webhook). Deliveries to it stop at once. Confirm the id with the user before calling — this is destructive.

• find_agents

  Search for agents you can reach by a fuzzy query (matched case-insensitively against name, handle, and description). Returns the union of agents you can access (your own + those shared through your orgs' projects) AND matching PUBLIC…

• generate_session_id

  Mint a new MCP session id bound to a specific agent. Required before any build_agent or chat_agent call. Pass the Agent Builder UUID ("b4409947-a46d-497e-94bb-fbba525185b7") to start a build_agent session; pass any other AgentSpec UUID to…

• get_agent

  Fetch one AgentSpec by id with every field: name, description, instructions, soul, chat_enabled / tool_enabled, system_tools, connector_ids, skill_ids, agent_tools, public_name, url, and timestamps. url is a direct link to open the agent…

• get_agent_reply

  Poll a chat_agent turn that returned status="working". Pass the same session_id; returns status="working", "complete", or "failed" with the final reply or user-safe failure text. Reports the session's LATEST turn — do not send another…

• list_agents

  List every agent you can access — your own plus those shared through your orgs' projects — as slim summaries (id, name, description, updated_at), newest first by updated_at. No instructions, soul, or tool lists; call get_agent for the…

• list_connectors

  List every ConnectorSpec the authenticated caller owns. Returns a slim summary (id, name, base_url, last_refreshed_at), newest first by last_refreshed_at.

• list_sessions

  List every MCP session the authenticated caller owns. Pass agent_id to scope the result to one agent; omit to return every session across every agent. Each entry carries the session_id, the bound agent_id, the display name, and the mint…

• list_webhooks

  List your registered webhooks with their event kinds, status (active or disabled), last delivery, last success, last error and failure streak. Secrets are never returned; delete and re-register an endpoint to get a new one.

• register_webhook

  Register a webhook: an https URL that receives a signed POST (Standard Webhooks headers webhook-id, webhook-timestamp, webhook-signature) whenever one of your conversations finishes a turn (conversation.reply), a scheduled action or…

• search_documentation

  Search the public Wajo documentation and get back the most relevant sections. No account or authentication required — use this to learn how to connect to Wajo, build and configure agents, use connectors and MCP servers, and drive the…

• test_webhook

  Send a webhook.test event to one of your webhooks now and report the HTTP status it answered, the latency and any error. Use it right after register_webhook to prove the endpoint is reachable and verifies signatures; a successful test…

• update_agent

  Patch an existing AgentSpec. Loads the current row, applies any fields you supplied (singular fields use proto presence — omit to leave alone), and saves. For instructions, prefer instructions_edits with a short verbatim old_text excerpt…