Flow Docs
Open Flow AI

MCP

Connect your AI tools — like Claude and Cursor — directly to Flow AI. Through the Model Context Protocol (MCP), your AI can search leads, build and manage lists, and message your LinkedIn network through Flow, on your behalf. This guide walks through connecting Claude, and the same steps apply to any MCP client.

How it works

MCP (the Model Context Protocol) is an open standard that lets AI apps securely use external tools. Flow AI runs an MCP server, so any MCP client — Claude, Cursor, and others — can call Flow’s tools: search for leads, create and manage lists, read and send messages, and pull pipeline stats. You stay in control: the connection uses an API key you create, the AI acts only through your Flow account, and anything outbound asks for your confirmation first.

Connecting takes three steps — create an API key in Flow, add Flow as a connector in your AI tool, then start prompting. Everything you need lives on one screen in Flow.

Flow's Settings → Connect with Claude screen: the MCP endpoint, a ready-to-paste config snippet, and an API keys table
Settings → Connect with Claude in Flow: your MCP endpoint, a ready-to-paste config snippet, and your API keys — all in one place.

Step 1 — Create an API key in Flow

In Flow, open Settings → Connect with Claude (app.chooseflow.ai/settings/mcp). In the API keys section, click Create key, give it a name (e.g. “Claude”), and copy it — the full key is shown only once. Creating keys is limited to company admins.

While you’re here, note the two things you’ll paste next: the MCP endpoint (https://api.chooseflow.ai/mcp) and the Claude config snippet, which already contains the endpoint with a placeholder for your key.

The API keys section in Flow with the Create key button highlighted
Create an API key (admins only). It replaces flow_live_YOUR_API_KEY in the config.

Step 2 — Add Flow to Claude

In Claude, open Settings → Connectors and choose Add custom connector.

Claude's Connectors settings with the Add custom connector button highlighted
In Claude’s Connectors settings, choose Add custom connector.

Give it a name (e.g. “Flow”) and paste your Flow MCP server URL with your key appended:

https://api.chooseflow.ai/mcp?token=flow_live_YOUR_API_KEY

Click Add. Flow’s tools are now available to Claude — no OAuth or extra setup needed.

Claude's Add custom connector dialog with Flow's MCP server URL pasted in and the Add button highlighted
Paste your Flow URL (with your key) into Remote MCP server URL and click Add.

Other clients — Cursor, Claude Desktop, config files

Every MCP client connects the same way. For tools that use a config file, paste the Claude config snippet from Flow (the mcpServers block) into the client’s MCP config:

  • Cursor — Settings → MCP → add a server, or edit ~/.cursor/mcp.json.
  • Claude Desktop — Settings → Developer → Edit config (claude_desktop_config.json).
  • Claude Code — add it to your project’s .mcp.json.
A code editor showing the Flow mcpServers config in ~/.cursor/mcp.json
The same mcpServers snippet drops straight into ~/.cursor/mcp.json — or any client’s config.

Remember to replace flow_live_YOUR_API_KEY with a real key in every client.

Start prompting

Once connected, just ask — Claude will call Flow’s tools when they’re relevant. For example:

  • “Search LinkedIn for heads of marketing at Series A SaaS companies and add the best 20 to a new list.”
  • “Show me connections in Follow Up who haven’t replied in a week, and draft a nudge for each.”
  • “What’s my connection acceptance rate this month?”

Anything that sends a message or starts a campaign asks for your confirmation first — the AI drafts and proposes, you approve.

Related

The API and Webhooks are other ways to integrate. MCP is the fastest way to put Flow in the hands of an AI assistant.

⌘I