MCP Server Overview

The FL Repos Intelligence MCP Server exposes the knowledge graph as a set of tools that any MCP-compatible AI agent can call. This enables conversational, natural-language access to the graph without writing API calls or Cypher queries manually.

What is MCP?

Model Context Protocol (MCP) is an open standard from Anthropic for connecting AI models to external data sources and tools. Any MCP host (Claude Desktop, Cursor, Windsurf, etc.) can connect to an MCP server and expose its tools to the AI model during a conversation.

Transports

The FL Repos MCP server supports two transports:

stdio

For Claude Desktop and Cursor. The host launches the server as a subprocess and communicates over stdin/stdout. Lowest latency, runs locally.

SSE (Server-Sent Events)

For network clients. Connect to the deployed server at https://fl-mcp-production.up.railway.app. No local setup required.

Claude Desktop Setup

Step 1: Locate the config file

~/Library/Application Support/Claude/claude_desktop_config.json

Create the file if it does not exist.

Step 2: Add the FL Repos server

{
  "mcpServers": {
    "fl-repos": {
      "command": "python",
      "args": ["-m", "intelligence.mcp.server"],
      "env": {
        "FL_API_URL": "https://api.agentic.forwardlane.com",
        "FL_API_TOKEN": "YOUR_API_TOKEN_HERE"
      }
    }
  }
}

Step 3: Restart Claude Desktop

After saving the config, quit and relaunch Claude Desktop. The FL Repos tools will appear in the tool list.

Step 4: Try a query

You: Which ForwardLane repositories have the most security vulnerabilities?
Claude: [calls list_security, summarizes findings]

Cursor Setup

Edit `.cursor/mcp.json`

{
  "fl-repos": {
    "command": "python",
    "args": ["-m", "intelligence.mcp.server"],
    "env": {
      "FL_API_URL": "https://api.agentic.forwardlane.com",
      "FL_API_TOKEN": "YOUR_API_TOKEN_HERE"
    }
  }
}

The FL Repos tools are now available in Cursor’s AI chat.

SSE Transport (Network Clients)

For clients that support SSE-based MCP connections, connect to:

https://fl-mcp-production.up.railway.app

The server at this URL already has the API token configured in its deployment environment. No additional authentication is required from the SSE client.

Example: Connecting with the MCP Python SDK

from mcp import ClientSession
from mcp.client.sse import sse_client

async with sse_client("https://fl-mcp-production.up.railway.app/sse") as (read, write):
    async with ClientSession(read, write) as session:
        await session.initialize()

        # List available tools
        tools = await session.list_tools()
        print([t.name for t in tools.tools])

        # Call a tool
        result = await session.call_tool(
            "search_repos",
            {"query": "authentication", "tier": 1}
        )
        print(result.content)

Read-Only vs. Write Mode

The MCP server starts in read-only mode by default. This means only the read tools are available (10 tools). Write tools are disabled to prevent AI agents from accidentally mutating the graph. To enable write tools, start the server with the --write flag:

python -m intelligence.mcp.server --write

Only enable write mode if you intentionally want the AI agent to be able to modify the graph. Write operations include adding nodes, updating properties, adding edges, and deleting nodes. Deletions are irreversible without a graph backup.

Environment Variables

Variable	Required	Description
`FL_API_URL`	Yes	REST API base URL.
`FL_API_TOKEN`	Yes	Bearer token for API authentication.
`GRAPHITI_URL`	No	Graphiti temporal memory server URL.
`ZEP_API_URL`	No	Zep episodic memory server URL.
`ZEP_API_KEY`	No	Zep API key (if Zep is configured).

Available Tools Summary

Tool	Mode	Description
`discover`	Read	Get graph schema, stats, and example queries.
`query`	Read	Execute a read-only Cypher query.
`search_repos`	Read	Search and filter repositories.
`get_repo_detail`	Read	Get full detail for one repository.
`get_recommendations`	Read	Get tier-based repo recommendations.
`list_functions`	Read	List functions from deep insights.
`list_models`	Read	List Django models.
`list_contributors`	Read	List contributor activity.
`list_security`	Read	List security pattern findings.
`search_memory`	Read	Search Graphiti/Zep memory.
`add_entity`	Write	Add a node to the graph.
`update_entity`	Write	Update properties on a node.
`add_relationship`	Write	Add an edge between two nodes.
`delete_entity`	Write	Delete a node from the graph.
`add_memory`	Write	Add an episode to Graphiti/Zep.

See MCP Tools for full parameter documentation for each tool.

Get Started

Products

API Reference

CLI

FL Web

Python SDK

MCP Server

Graph Schema

MCP Server Overview

MCP Server Overview

What is MCP?

Transports

stdio

SSE (Server-Sent Events)

Claude Desktop Setup

Step 1: Locate the config file

Step 2: Add the FL Repos server

Step 3: Restart Claude Desktop

Step 4: Try a query

Cursor Setup

Edit `.cursor/mcp.json`

SSE Transport (Network Clients)

Example: Connecting with the MCP Python SDK

Read-Only vs. Write Mode

Environment Variables

Available Tools Summary

​MCP Server Overview

​What is MCP?

​Transports

stdio

SSE (Server-Sent Events)

​Claude Desktop Setup

​Step 1: Locate the config file

​Step 2: Add the FL Repos server

​Step 3: Restart Claude Desktop

​Step 4: Try a query

​Cursor Setup

​Edit .cursor/mcp.json

​SSE Transport (Network Clients)

​Example: Connecting with the MCP Python SDK

​Read-Only vs. Write Mode

​Environment Variables

​Available Tools Summary

MCP Server Overview

What is MCP?

Transports

Claude Desktop Setup

Step 1: Locate the config file

Step 2: Add the FL Repos server

Step 3: Restart Claude Desktop

Step 4: Try a query

Cursor Setup

Edit `.cursor/mcp.json`

SSE Transport (Network Clients)

Example: Connecting with the MCP Python SDK

Read-Only vs. Write Mode

Environment Variables

Available Tools Summary