MCP Server

GitDealFlow — MCP Server

Track startup engineering acceleration from public GitHub data. Discover high-momentum startups before they announce funding.

How the signal works: GitDealFlow reads public GitHub activity across 4,200+ startups and tracks commit velocity, contributor growth, and repository expansion. When those accelerate together, the team is typically gearing up to raise — in a research panel of 219 documented fundraises, the acceleration led the public round announcement by 21-47 days. The weekly email delivers five names every Sunday, and this MCP server exposes the same dataset to agents.

The GitDealFlow MCP server exposes the startup momentum dataset to any MCP-compatible client. All tools are read-only, idempotent, and completely free — no authentication, no API key, no rate-limit gating. The data is derived from public GitHub activity across 4,200+ startups, with the methodology published on SSRN under CC BY 4.0 alongside an open dataset.

Quick start (stdio)

For Claude Desktop, Claude Code, Cursor, or any client that supports local MCP servers, run the published package directly:

npx -y @gitdealflow/mcp-signal

Streamable HTTP transport

For hosted MCP clients (ChatGPT apps, remote MCP runtimes) that prefer a URL endpoint, use the Streamable HTTP server:

https://signals.gitdealflow.com/api/mcp/rpc

Install in Claude Desktop

Add to claude_desktop_config.json:

{
 "mcpServers": {
   "gitdealflow": {
     "command": "npx",
     "args": ["-y", "@gitdealflow/mcp-signal"]
   }
 }
}

Install in Cursor

Add to .cursor/mcp.json:

{
 "mcpServers": {
   "gitdealflow": {
     "command": "npx",
     "args": ["-y", "@gitdealflow/mcp-signal"]
   }
 }
}

Inspect

npx @modelcontextprotocol/inspector npx -y @gitdealflow/mcp-signal

Available Tools (6)

get_trending_startups — Top 20 accelerating startups across all sectors.
search_startups_by_sector — Filter the tracked universe by sector (20 enumerated sector slugs).
get_startup_signal — Engineering momentum signal for a specific startup; case-insensitive and normalization-tolerant.
get_signals_summary — Period, freshness, and format URLs for the current signal dataset.
get_scout_receipts — Scout Score (0-100) for a GitHub user, computed from their starring history against validated breakout companies.
get_methodology — Full methodology text and canonical URL, so any agent can verify how the signal is built.

Distribution and discovery

npm: @gitdealflow/mcp-signal
MCP Registry: io.github.kindrat86/vc-deal-flow-signal
Glama: A-tier listing — glama.ai/mcp/servers/@kindrat86/vc-deal-flow-signal
Discovery manifest: /.well-known/mcp.json

Other programmatic surfaces

Prefer plain HTTP? The same dataset is available without any MCP client: JSON API at https://signals.gitdealflow.com/api/signals.json, CSV export at /api/signals.csv, a function-calling API (/api/agent/tools + /api/agent/call) for OpenAI/Anthropic/Gemini SDKs, an A2A endpoint at /api/a2a, an NLWeb endpoint at /api/nlweb, and a full OpenAPI 3.1 spec at /api/openapi.json. All free and unauthenticated.

← Back to GitDealFlow