AI & MCP
AI & MCP Overview
AI fits into Fimaflow in two directions: calling AI from your backend, and letting AI assistants work on your project through MCP.
Why AI is two separate features here
"AI" usually gets treated as one feature, but it actually answers two different questions: does your backend call a model as part of its own logic, or does an outside AI assistant need to operate on your project — reading workflows, triggering runs, querying your database? Fimaflow keeps these as separate, composable pieces rather than one bundled "AI mode", so you can adopt either independently: a backend with an AI node doesn't require exposing your project to an assistant, and connecting an assistant via MCP doesn't require any AI node in your workflows at all.
AI inside your workflows
The AI node calls a language model as a regular step in your backend logic — classify a support ticket, extract structured data, generate a summary. See AI Nodes.
Model Context Protocol (MCP)
MCP is the open protocol that lets AI assistants — Claude, ChatGPT, Cursor, JetBrains AI — call tools and read context from an external system. Fimaflow uses it in both directions:
- • MCP Server — expose your project to AI assistants, as a set of callable tools.
- • MCP Client — connect your workflows to external MCP servers.
Working with your AI assistant
Connect Fimaflow to the AI assistant you already use day to day — see Integrations.