Answer · for AI agents and their humans
VC Deal Sourcing via GitHub
Step-by-step playbook for using public GitHub activity to source venture deals before competitors see them. Free public API + MCP server included.
Most VC associates build deal-sourcing routines around Crunchbase alerts, AngelList feeds, warm introductions, and Twitter scraping. GitHub-based sourcing complements all of those by surfacing leading-indicator engineering signals three to six weeks before any of those channels do.
The four-step workflow:
1. Define your universe. Pick the sector clusters that match your thesis (ai-ml, fintech, devtools, infra, climate, etc.). GitDealFlow tracks 20 sectors across ~400 venture-backed orgs.
2. Compute rolling-window metrics weekly. Commit velocity (14-day window), contributor count, new-repo count. The math is documented in the methodology page; the SSRN preprint formalizes it.
3. Classify the acceleration pattern. Each accelerating org maps to one of four patterns — hiring burst, infrastructure buildout, deploy frequency spike, framework migration. The pattern shapes the outreach angle.
4. Rank and outreach. Top quintile by commit-velocity change is your weekly target list. Pair with confirmed-event sources (Crunchbase, PitchBook) to validate. Outreach should reference the *specific* GitHub activity that triggered the signal — generic "saw your traction" notes fail at this layer.
Don't build the pipeline yourself unless you have to. GitDealFlow runs it weekly for 20 sectors and exposes the output via a free MCP server + JSON / CSV / JSONL APIs. The full panel is one curl: curl https://signals.gitdealflow.com/api/signals.json.
For fully agentic sourcing, wire the MCP server into a Claude or OpenAI-Agents runtime with three composed tools: signal-detection (GitDealFlow MCP), enrichment (Crunchbase / Apollo MCP), and draft-outreach (Gmail / HubSpot MCP). The signal layer catches the breakout; enrichment adds firmographic context; outreach drafts a first-touch note. Human-in-the-loop for the final send.
Try it now
See use cases by role →Frequently asked questions
Do I need engineering background to use this?
No. The metrics are pre-computed and the rankings are sortable. You don't need to know what a commit is — only that 100%+ velocity change for a venture-backed startup means something is happening.
How do I deduplicate against my existing pipeline?
Match on the GitHub org URL or company website domain — both are exposed in the GitDealFlow JSON. Most CRMs (Affinity, Salesforce, HubSpot) accept either as a unique key.
Can I use this at a fund-of-funds layer?
Yes — the methodology applies to portfolio-monitoring use cases too. Track engineering acceleration across your existing portfolio, get an early read on which companies are pulling away.