MCP Server
The Frontrun MCP server gives AI agents native access to the Frontrun API. It exposes all API endpoints as MCP tools that agents can discover and call directly.
No local setup required — everything runs via npx.
Setup
Step 1: Get your API key
Go to frontrun.vc (opens in a new tab) → Settings → API Keys and create a key.
Step 2: Connect to your client
Choose your client and follow the instructions below.
Claude Code
Run this single command in your terminal:
claude mcp add frontrun -e FRONTRUN_API_KEY=your_api_key --scope user -- npx frontrun-mcp-serverThat's it. The --scope user flag makes it available in every project. Verify it's connected:
claude mcp listYou should see frontrun in the list. Start Claude Code and ask: "What's trending in VC follows this week?"
Claude Desktop
Add to your config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"frontrun": {
"command": "npx",
"args": ["frontrun-mcp-server"],
"env": {
"FRONTRUN_API_KEY": "your_api_key"
}
}
}
}Restart Claude Desktop. The Frontrun tools will appear automatically.
Cursor
Add to .cursor/mcp.json in your project root:
{
"mcpServers": {
"frontrun": {
"command": "npx",
"args": ["frontrun-mcp-server"],
"env": {
"FRONTRUN_API_KEY": "your_api_key"
}
}
}
}Python
import subprocess, json
process = subprocess.Popen(
["npx", "frontrun-mcp-server"],
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
env={"FRONTRUN_API_KEY": "your_api_key", "PATH": "/usr/local/bin:/usr/bin"}
)JavaScript
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
const transport = new StdioClientTransport({
command: "npx",
args: ["frontrun-mcp-server"],
env: { FRONTRUN_API_KEY: "your_api_key" },
});
const client = new Client({ name: "my-app", version: "1.0.0" });
await client.connect(transport);Available tools
| Tool | Description |
|---|---|
frontrun_status | Account status, balance, usage stats |
frontrun_list_tracked | List all monitored accounts |
frontrun_track | Start monitoring an X account |
frontrun_untrack | Stop monitoring an X account |
frontrun_new_follows | Detect new follows across tracked accounts |
frontrun_snapshot | Get current follow list for an account |
frontrun_convergence | Detect multi-account convergence signals |
frontrun_trending | Get trending entities by follower count |
frontrun_account_activity | Activity profile: velocity, sectors, recent follows |
frontrun_search | Search entities by sector, keyword, or type |
frontrun_enriched_follows | New follows with full enrichment |
frontrun_classify | Run classification on specific entities |
frontrun_create_rule | Create custom classification rules |
frontrun_list_rules | List classification rules |
frontrun_update_rule | Update a classification rule |
frontrun_delete_rule | Delete a classification rule |
frontrun_tag | Add custom tags/notes to entities |
frontrun_list_tags | List your custom-tagged entities |
Example prompts
Once connected, ask your agent:
- "What are the trending companies this week?"
- "Show me convergence signals with threshold 3 in the last 14 days"
- "What new accounts did pmarca follow in the last 48 hours?"
- "Search for AI/ML startups in the follow graph"
- "Track @sequoia"
- "Tag @openai as a competitor with sector AI"
Troubleshooting
"FRONTRUN_API_KEY environment variable is required" Your API key isn't set. Double-check the key in your config or command.
"Invalid API key" The key is wrong or inactive. Generate a new one at frontrun.vc → Settings → API Keys.
npx command not found Install Node.js 18+ from nodejs.org (opens in a new tab).
Compatible clients
Any MCP-compatible client works:
- Claude Code (CLI)
- Claude Desktop
- Cursor
- Windsurf
- Continue
- Custom MCP clients
Source code
The MCP server is open source: github.com/jongall45/frontrun-mcp-server (opens in a new tab)