Flow Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open standard designed to streamline how AI models connect to external data sources and local tools. It replaces the need for custom, one-off integrations by providing a universal plug-and-play interface between AI applications and various data ecosystems.

By using MCP, developers can grant models secure, standardized access to everything from local files and databases to third-party services like Flow. This creates a more cohesive environment where AI agents can interact with real-time information across different platforms without specialized re-coding.

Follow this guide to learn how to configure MCP clients to connect to Flow's MCP server.

Prerequisites

Create a Flow API key if you don't already have one.

Endpoint

https://appfireflow.com/ai/api/v1/mcp/

Client configuration

Cursor

Add the following to your Cursor MCP settings (.cursor/mcp.json or via Settings > Cursor Settings > Tools & MCP):

{
  "mcpServers": {
    "flow-mcp": {
      "url": "https://appfireflow.com/ai/api/v1/mcp/",
      "headers": {
        "Authorization": "Bearer <Flow-API-Key>"
      }
    }
  }
}

Claude Code

Run the following command:

claude mcp add flow-mcp --transport http https://appfireflow.com/ai/api/v1/mcp/ --header 'Authorization: Bearer <Flow-API-Key>'

Available tools

Once connected, your MCP client has access to Flow tools, including:

Team and user data - List teams, team members, and users.
Git metrics - Commit counts, coding days, PR statistics.
Ticket metrics - Sprint data, delivery metrics, cycle time.
DORA metrics - Deployment frequency, Lead time, and Change failure rate.
Flow documentation search - Search Flow's help documentation.

Troubleshooting

Authentication errors

Verify your API key is valid and not expired.
Ensure the Authorization header uses the Bearer prefix.
The Flow MCP server may briefly show a red dot in Cursor's MCP settings after initial setup. This is a cosmetic issue — the connection is working and all tools are available. If you see a red indicator, try restarting Cursor and it should turn green.

Permission errors

You can only access data you have permission to view in Flow.
Contact your Flow admin if you need additional permissions.

Example: Advanced portfolio visualization & strategic analysis

The MCP does not just put Flow data in a new window; it unlocks the data. It allows leaders to use the specialized UI components of tools like Canvas—which Flow does not natively possess—to transform raw engineering metrics into a compelling strategic narrative.

Scenario

As an Engineering Director, you need to present the quarterly performance review (Q4 2025 vs. Q1 2026) to stakeholders. While the Flow AI Assistant provides raw data and basic insights, you need high-fidelity, side-by-side visualizations and a long-form narrative report that correlates git activity with sprint predictability.

Solving it using Flow + other AI tools

By accessing the Flow AI Assistant, for example, through Canvas, you can go beyond simple chat responses. Thanks to the MCP, you can seamlessly pull real-time data from Flow while leveraging the advanced computational and creative capabilities of the host AI tool.

Requested Flow data received and visualized in Canvas.

Key benefits

Dynamic visualizations: While the native Flow Assistant is optimized for text-based insights, pulling that same data into Canvas enables the AI to generate complex, presentation-ready bar charts (for example, comparing Throughput and Cycle time across quarters).
Multi-source correlation: You can ask the AI to cross-check PR behavior with ticket trends. The AI uses Flow’s data to identify that.
Automated executive summaries: The AI tool leverages its long context window to draft a comprehensive bottom-line assessment, highlighting exactly where the red flags are appearing (for example, Queue Time trending worse) and providing actionable recommendations for the next quarter.