Skip to content

hebbs-ai/vcbrain

Repository files navigation

VCBrain — the open-source AI analyst for venture capital

Catch every deal. Research it against your thesis. Draft the IC memo. Watch the portfolio. Self-hosted, agent-native, MIT-licensed.

VCBrain is an open-source AI deal team you run yourself. It catches startup leads from every channel, turns each one into an analyst-grade dossier, scores it against your fund's actual thesis, drafts the IC memo, and keeps watch on your portfolio — so a lean team can run the dealflow of a much bigger one without anything slipping through.

Built as a module on the BoringOS agent framework. Your dealflow and data stay on your infrastructure.

Want this dialed in for your fund — your thesis, your sources, a managed deployment? Talk to us at hebbs.ai.


Why VCs run VCBrain

  • Nothing drops. Inbound email, an embeddable submission form, your copilot chat, and an automated scout all funnel into one deduped pipeline.
  • Thesis-aware, not generic. Most tools score startups on vanity signals. VCBrain scores against the thesis you configure — hard must-haves, deal-breakers, weighted dimensions — and shows its work.
  • Analyst-grade dossiers. Reads the pitch deck, researches the founders, market, competition, traction (with unit economics) and the cap table, then writes a sourced brief with a diligence plan — every claim cited.
  • Open source + self-hostable. No black box, no data leaving your stack. MIT-licensed: fork it, extend it, run it commercially.
  • Agent-native. Real autonomous agents do the work on a schedule and on every new lead — not a chatbot you have to babysit.

How it works

  Email   Public form   Copilot   Scout
     \________\___________/_______/
                  │  (deduped intake — one record per company)
                  ▼
          vc-research  ──►  living dossier (deck + deep web research)
                  │
                  ▼
        vc-thesis-fit  ──►  0–100 score vs your thesis, fits / risks / conviction
                  │
                  ▼
        vc-memo-writer ──►  cited IC memo, ready to edit & decide
                  │
                  ▼
  vc-portfolio-monitor  ──►  daily signals on invested companies
        vc-scout        ──►  daily thesis-fit cold leads (YC, Product Hunt, GitHub)

The five agents

Agent What it does Trigger
vc-research Builds the living dossier — deck + web research, founders, market, competition, traction, cap table, diligence plan New lead (any channel)
vc-thesis-fit Scores 0–100 against your active thesis; fits, risks, conviction, comps, return scenarios Lead reaches Screening
vc-memo-writer Drafts a cited IC memo from the dossier + score Lead reaches IC
vc-portfolio-monitor Records news / hiring / runway / KPI signals on invested companies Daily cron
vc-scout Surfaces thesis-fit cold leads from public sources Daily cron

The interface

  • Pipeline — kanban by stage, fit score on every card.
  • Dossier — the analyst brief: founder deep-dives, unit economics, competition, a diligence plan, live signals, the IC memo, and every source.
  • Thesis Studio — define must-haves, deal-breakers and dimension weights; back-test against recent submissions.

Install

VCBrain is a BoringOS module — install it into a running BoringOS host (no separate server to operate).

pnpm install
pnpm build          # builds server + web (the PluginUI bundle)
pnpm test:run       # 27 tests

Then pack and upload the .hebbsmod to your host's Apps screen, or via the admin API. See docs/PLAN.md for architecture and docs/DECISIONS.md for the full build log.

Requires @boringos/connector-google >= 0.2.13 on the host for Gmail attachment download (pitch decks).

Repo layout

packages/server   # the module: tools, 5 agents (SKILL.md), schema, lifecycle, webhook
packages/web      # PluginUI: Pipeline, Dossier, Thesis Studio, dashboard widgets
packages/shared   # shared DTOs + constants
docs/             # PLAN.md (architecture) + DECISIONS.md (build log)

License

MIT — see LICENSE. Free to use, fork, self-host, and build on commercially.


Built by Hebbs. We build agentic operating layers for high-leverage teams. Running a fund and want this tailored to your thesis and sources? → hebbs.ai

About

Open-source AI analyst for VCs — catch every startup lead, research it into an analyst-grade dossier, score it against your thesis, and draft the IC memo. Self-hosted, agent-native, MIT. Built on BoringOS. → hebbs.ai

Topics

Resources

License

MIT, Unknown licenses found

Licenses found

MIT
LICENSE
Unknown
LICENSE.md

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors