AI Agents and MCP

Talk to your integration: the AI agent investigates, and only acts after your confirmation.

The Fluid MCP Server connects AI agents, such as Claude and Cursor, to your Fluid environment. Instead of opening the console and filtering screens, you ask in natural language what failed, the agent shows you where, and reprocesses the event if you authorize it.

Talk to a specialist →View the MCP Server documentation

Fluid + MCP

From the systems you already have to AI agents

The integrations Fluid already orchestrates become tools of an MCP Server. Agents like Claude, ChatGPT or Cursor can use ERPs, CRMs and APIs with the same governance as your flows.

  1. ERP / CRM / APIs
  2. Fluid Flow
  3. MCP Tool
  4. AI Agent

In practice

A real usage example

Friday night, a failure alert, nobody from the technical team available. The operator asks the agent whether there were failures today, requests the details of the failed execution, sees which step it stopped at, and authorizes reprocessing the event. All in one conversation, without opening the console.

Fluid MCP Server

What the agent does for you

  • Monitor

    Lists executions by status, source, period, and tag.

  • Investigate

    Shows an execution's status, the step where it failed, and what was sent and returned at each step.

  • Reprocess, with confirmation

    Triggers a new execution of an event, with the original or adjusted data.

  • Track FlowKits

    Shows which are active or failing and activates a new one, also with confirmation.

Who it is for

Operations and support teams that need to act on failures without waiting for the technical team. Customer Success tracking a customer's installation without opening the console. Managers who want to see executions without a technical tool. Developers who already work with Claude or Cursor.

Security

Control and security

  • Every request is isolated by workspace: you only see data from your own environment.

  • Write operations always require explicit confirmation before running.

  • Your credentials never appear in the agent's responses.

  • Access uses an MCP-scoped credential and a token with an expiration you define.

How to connect

Three steps: create an MCP-scoped credential in the console, generate an access token in the Fluid Playground, and register the server in your MCP client. Works with Claude Desktop, Claude Code, Cursor, and other compatible clients.

Model Context Protocol

What is MCP

The Model Context Protocol is an open standard that lets AI agents connect to external tools and act on the user's behalf. It works as a standard fit between the agent and the systems you already use.

AI in flow building too

Monacopilot suggests code and data transformations right in the flow editor. And the ChatGPT (OpenAI) connector can be used in flows like any other connector.

FAQ

Frequently asked questions

Can the agent change anything without my authorization?

No. Reprocessing an event and activating a FlowKit always require explicit confirmation.

Can I see data from other customers or workspaces?

No. Every request is isolated by workspace.

Which agents work with the Fluid MCP Server?

Claude Desktop, Claude Code, Cursor, and other MCP-compatible clients.

Next time a flow fails, the first answer can come from a question, not from searching through screens.

Talk to a specialist →View the MCP Server documentation