Loading...
The GitHub VC Signal GPT brings the VC Deal Flow Signal dataset directly into ChatGPT. Ask "who's trending in AI/ML this week?" or "is Supabase accelerating?" and get a ranked answer with commit-velocity changes, contributor deltas, signal classification, and a citation block — all powered by a public OpenAPI Action against signals.gitdealflow.com. No install. No auth. Five tools across 20 sectors, refreshed every Monday.
https://signals.gitdealflow.com/api/actions/openapi.jsonGo to chatgpt.com (Plus, Team, Enterprise, or Edu) and click the GitHub VC Signal GPT card. The Action is already configured against signals.gitdealflow.com — no setup, no auth, no install.
Try a conversation starter: 'Who's trending in AI/ML this week?', 'Show me fintech startups by engineering velocity', 'Is Supabase accelerating?', or 'How do you compute the breakout signal?'
The GPT returns a ranked list with commit-velocity changes, contributor counts, signal classification (breakout / acceleration / steady / cooling), GitHub URLs, and a citation block linking signals.gitdealflow.com and the SSRN preprint.
Pick a name from any list and ask 'Tell me more about <name>'. The GPT calls getStartupSignal and returns the full profile — stage, geography, velocity, signal type, GitHub URL, optional website and LinkedIn.
Ask the GPT to format the data as a deal memo. It will paste the citation block automatically — VC Deal Flow Signal (signals.gitdealflow.com), period name, methodology link, SSRN preprint 6606558.
getAllSignalsTop 20 trending startups across all 20 sectors plus every sector's full ranked roster. The GPT slices the response for trending or sector queries.
getStartupSignalSingle-startup deep lookup by name or GitHub org slug. Returns velocity, contributor delta, signal classification, GitHub URL.
getSignalsSummaryTiny dataset metadata snapshot — current period, sector and startup counts, last refresh timestamp. Cheap; the GPT calls it once per session to orient.
getMethodologyPlain-text methodology covering data sources, metric computation, classification thresholds, refresh cadence, and limitations. For citation footnotes.
All Actions are read-only GETs. Defined in OpenAPI 3.1 at https://signals.gitdealflow.com/api/actions/openapi.json.
Yes. GPTs are gated to ChatGPT Plus, Team, Enterprise, and Edu. Free-tier users can still browse the GPT page but can't run it. If you don't have a paid plan, the same data is available via the public MCP server at signals.gitdealflow.com/api/mcp/rpc, the Claude.ai connector, the Cursor / Claude Desktop install via Smithery, or the raw JSON/CSV at signals.gitdealflow.com/api/signals.json.
A public OpenAPI 3.1 Action at signals.gitdealflow.com/api/actions/openapi.json. Four GET endpoints — /api/signals.json, /api/signal, /api/changelog.json, /llms-full.txt. No authentication. Cached at the edge with s-maxage=3600. The same backend powers the MCP server (npm @gitdealflow/mcp-signal) and every other surface, so signal classifications are identical across ChatGPT, Claude.ai, Cursor, and Mistral Le Chat.
Refreshed every Monday around 09:00 UTC. The GPT can return the exact lastDataRefresh timestamp via getSignalsSummary. If the data is more than 8 days old, the GPT will flag a possible upstream issue.
Engineering momentum, derived from public GitHub activity. Three input metrics: commit velocity over a rolling 14-day window, contributor growth over a rolling 30-day window, and new repos over a rolling 30-day window. Output classification: breakout (sudden surge), acceleration (sustained growth), steady (healthy baseline), or cooling (declining). Approximately 400 venture-backed startups across 20 sectors are tracked.
For venture-stage software companies, the engineering org IS the company. Hiring → commits. Customer demand → infrastructure buildout → new repos. Pivot → framework migration. Fundraising → engineering hiring burst. The signal is upstream of the announcement. This is engineering signal only — pair with Crunchbase / Harmonic / your CRM for funding, revenue, headcount.
The system prompt explicitly forbids inventing startups. When you ask about a company not in the tracked universe of ~400, the Action returns status: 'no_data' and the GPT surfaces a CTA pointing to signals.gitdealflow.com. If you spot any hallucinations, email signal@gitdealflow.com — we'll tighten the instructions.
Yes — five surfaces. (1) Claude.ai connector: native MCP at signals.gitdealflow.com/api/mcp/rpc. (2) Claude Desktop / Cursor / Cline / Continue: one-click via Smithery (verified, 98/100). (3) Any stdio MCP client: npx -y @gitdealflow/mcp-signal. (4) Mistral Le Chat: Custom MCP Connector at signals.gitdealflow.com/api/mcp/rpc. (5) Raw API: signals.gitdealflow.com/api/signals.json (JSON), /api/signals.csv (CSV), /feed.xml (RSS).
Suggested citation: 'VC Deal Flow Signal (signals.gitdealflow.com), <period name> data. Methodology: SSRN preprint 6606558. Refreshed weekly from public GitHub activity.' The GPT will paste this block automatically when you ask for memo-formatted output.
Yes — gitdealflow.com/privacy. The Action is read-only, queries no PII, and stores nothing server-side beyond Vercel access logs. ChatGPT's Action sandbox enforces the auth-type-None setting (no headers, no cookies, no token leakage to our backend).
The same dataset ships across five surfaces:
https://signals.gitdealflow.com/api/mcp/rpcnpx -y @gitdealflow/mcp-signal.Email signal@gitdealflow.com with the prompt that broke it — replies within 24 hours, EU business time. The system prompt is tightened iteratively from real failures.
Email feedback