Aleph
Query live FP&A data, build reports and update dashboards from your AI assistant.
Discover, govern and monitor every API, agent, MCP server and LLM in your Anypoint estate from an AI assistant.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The MuleSoft MCP server connects MuleSoft Platform to Claude, ChatGPT and any MCP-compatible agent. It exposes 66 tools that discover, govern, monitor and cost-analyse every API, agent, MCP server and LLM across an Anypoint estate. Sign-in is OAuth through Anypoint Platform, using a connected app you create yourself before connecting.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
full and offline_access, and MuleSoft's troubleshooting tells you to add Full Access and Background Access to the connected app when authorization fails. On a 66-tool server that can apply policies to live API instances and provision gateways, that is the single most important fact on this page. Anypoint's own authorization server publishes 133 scopes including read:full, so a narrower grant exists at the platform level — it is simply not what this MCP resource offers. Constrain access with Anypoint roles instead, not with the OAuth grant.control names in Anthropic's directory and guardrail names in MuleSoft's reference, and three further names appear on only one side. We could not resolve which the live server answers to.has_mcp_app: true in the snapshot of 2026-08-16), and MuleSoft's tool reference repeatedly describes tools that "launch a wizard" for gateway, scanner, policy and guardrail creation. We could not confirm what that interactive surface renders, because reading it requires a session we do not have.MuleSoft Platform MCP Server, the hosted one at omni.mulesoft.com. MuleSoft ships three MCP things and only this one is a remote connector. MuleSoft DX MCP Server is a local npm package for IDE development work, and the MuleSoft MCP Connector is a component you drop inside a Mule application to expose your own flows as tools.
Full Access, plus Background Access for token refresh. When we queried the endpoint's own protected-resource descriptor on 2026-08-18 it advertised exactly two scopes, full and offline_access. There is no read-only scope on offer, so a token that can read your governance reports can also apply policies.
Yes, and you have to create it yourself first. MuleSoft's setup guide requires a connected app that acts on the user's behalf, with the Authorization Code and Refresh Token grant types selected. Its documentation states the servers do not support Dynamic Client Registration, so the client ID and secret are manual.
Almost always a redirect URL mismatch. MuleSoft's troubleshooting explains that IDE clients use a localhost callback whose port changes each session, so the URL stops matching the one registered in your connected app. Fix it by pinning a fixed callback port and registering that exact localhost URL.
It depends on the tool, and MuleSoft publishes the mapping per tool. Reading APIs needs Exchange Viewer, monitoring needs Monitoring Viewer, applying a policy needs API Manager Environment Admin, and creating a guardrail needs Governance Administrator. Your existing Anypoint org roles govern what the assistant can do.
Yes. The tool list includes applying policies to live API instances, creating managed gateways and scanners, provisioning MCP servers on Flex Gateway, and deleting governance strategies. MuleSoft documents these as actions taken on your behalf, gated by Anypoint role rather than by any per-tool read-only switch.
https://omni.mulesoft.com/mcp — an anonymous initialize returns HTTP 401 with WWW-Authenticate: Bearer realm="mcp" pointing at an RFC 9728 protected-resource descriptor. That descriptor returns HTTP 200 and declares authorization_servers: ["https://anypoint.mulesoft.com"] with scopes_supported: ["full", "offline_access"]. This is our own observation on 2026-08-18, and it is why the sign-in requirement and the missing read-only scope are stated as verified rather than quoted. It is also why no tool descriptions, parameter schemas or safety annotations appear on this page: the handshake never reached tools/list. The /mcp/sse path and the eu1 regional host returned the same challenge on the same date. · retrieved 2026-08-18code_challenge_methods_supported: ["S256"]) · retrieved 2026-08-18docs.mulesoft.com/robots.txt allows crawling and carries Content-Signal: ai-input=yes, ai-train=no, search=yes — synthesis into AI-generated output is expressly permitted by the publisher, and this page honours the training refusal. · retrieved 2026-08-18has_mcp_app flag (from the committed directory snapshot of 2026-08-16) · retrieved 2026-08-16Connect MuleSoft once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.