The company brain for VC funds.

Funds run on knowledge locked in people's heads and data scattered across email, CRMs and pitch decks. VCbrain structures it into one living brain — so deal flow, due diligence and fund intelligence stop being manual work.

See the live demo
For 5–10 person funds
DACH · Europe
GDPR-native
Every company in the world is going to need one.
Garry Tan, YC President

The opportunity

Every company. Not just VCs.

01The macro shift

The models are no longer the blocker.

Yesterday

Models were the blocker

GPT-3 couldn't reason. Fine-tuning was expensive. Prompt engineering was an art. The AI wasn't good enough.

Today

Domain knowledge is the blocker

Models are extraordinary. But they can't operate on scattered, tacit, undocumented company knowledge. The AI is good enough. The data isn't.

← We are here

Tomorrow

Company brains unlock everything

Every company has a structured, current, executable knowledge layer. AI agents operate reliably on real company context. Every workflow automated.

02The problem

Four areas where small funds drown in manual work.

A typical fund of 2–10 professionals sees hundreds of deals a year, runs months-long DD, manages 10–20 portfolio companies and reports to LPs — with fragmented tooling. Ordered by how a fund experiences and pays for them.

01Entry point · easy to approve

Deal flow management

Deals arrive from everywhere — conferences, intros, forwarded decks, LinkedIn. A junior manually reads each one and types it into the CRM. Repetitive, error-prone, slow. Important deals sit in inboxes for days.

VCbrain. Inbound connects to VCbrain. A deck arrives → entities extracted, thesis precheck scored, a structured deal record created automatically. The team sees a clean, scored record — not a raw attachment.

02Biggest bottleneck · complex

Due diligence

DD means synthesising a data room, desk research, reference calls, email threads and transcripts into a long memo. An IM can spend 2–4 weeks on a single DD — most of it gathering and formatting, not judgment.

VCbrain. Every source feeds VCbrain. Parallel agents run commercial, legal, tech, market and competitive analysis into structured output. The human makes the judgment calls and writes the verdict. The memo writes itself from structured inputs.

03Strategic · high excitement

Fund intelligence

The most valuable knowledge — why they passed, what patterns they've seen, how the thesis evolved — lives in partners' heads and IC debate notes. When a partner leaves, it walks out the door.

VCbrain. Every IC decision, pass reason and reference-call outcome is structured into a fact graph and pattern library. Ask anything in plain language — "What's our current HealthTech exposure?" — backed by live data, not static documents.

04Unique to VCbrain

Deal source intelligence

Every fund has a gut feel that "conference X is good." Nobody has it quantified. Funds spend equal time on cold inbound (2% conversion) and warm angel referrals (33%) because they can't see the difference.

VCbrain. Because VCbrain ingests every deal with its source metadata, it scores each source by conversion — seen vs reached-IC vs invested. After 60 days the data exists; after 6 months the patterns are clear.

03The product

Six knowledge layers. One living brain.

Raw company data becomes a structured, current, provenance-tracked brain. A skills file turns it into executable knowledge. AI agents then automate reliably — not hallucinate on stale docs.

  1. 1

    Thesis layer

    Real beliefs, versioned. Confidence levels, disagreements, evidence base — not the LP-deck version.

  2. 2

    Pattern library

    What signals correlated with wins. What founder archetypes worked. The misses — passed deals that became huge.

  3. 3

    Deal memory

    Every company ever touched: why it came in, internal debate, decision made, reasoning — not just the outcome.

  4. 4

    Network graph

    Who knows who, relationship strength, domain expertise, who gives real founder signal.

  5. 5

    Portfolio intelligence

    Who solved this problem already? Which portfolio company should this founder talk to?

  6. 6

    Self-improving harness

    Every IC correction becomes a training signal. The brain gets smarter from your decisions.

Skills file output

  • First-pass evaluation
  • Reference identification
  • IC memo drafting
  • Thesis drift detection
  • Portfolio matching
  • LP report generation
  • Competitive synthesis
04Why VC first

VC funds are the perfect beachhead.

01

Pain is acute and costly

A wrong IC decision costs €500K+. A missed pattern costs a fund its vintage. The ROI of better judgment is enormous — and measurable.

02

Knowledge is unusually tacit

VC knowledge is almost entirely in heads. Why they passed. What patterns they've seen. How their thesis evolved. None of it is written down.

03

n8n penetration is high

Tech-forward funds already use n8n to automate workflows. They have the pipes. They just don't have the brain. We plug into existing infrastructure.

04

Network effects compound

Co-investor graphs create cross-fund value. A fund that shares deal flow benefits from shared pattern libraries. Winner-takes-most at the category level.

05

They understand the product

VCs invest in infrastructure primitives. They grasp 'company brain' immediately. First customers are also evangelists to their portfolio.

06

Beachhead to every company

Every VC-backed company will ask: 'can we have this too?' VCbrain becomes the wedge into the €100B+ enterprise knowledge management market.

05Data security

A first-class concern, not a feature.

VCs can't start an AI conversation without a credible answer to "where does my data go?" It must be solved and demoable before the first fund meeting — it determines whether the conversation continues at all.

EU-compliant API layer

Suggested default

Route all LLM calls through an EU-compliant API layer (e.g. Langdock). Data stays in EU data centres, GDPR-compliant, no model training on customer data.

Best for. Most EU-based funds. Fastest to deploy.

On-premise local LLM

For larger funds

Host an open-source model (Mistral, LLaMA) on the fund's own infrastructure. Zero data leaves their environment.

Best for. Larger funds with strict data policies. Higher setup cost.

Security is a gatekeeper, not a feature — answered before the first fund meeting, not after.

06Integration approach

Two ways to connect your stack.

Both deliver the same outcome — structured deal data flowing into VCbrain. The right choice depends on what your fund already runs and what it values.

n8n workflow

Deterministic · auditable

We wire your inbound to VCbrain as visible n8n nodes. A trigger fires, data is extracted, a router branches the logic, and a record is created — every step editable on the canvas. When something breaks, you see exactly which node.

Best for.
Funds already on n8n; teams that want full pipeline visibility.
Tradeoff.
Less flexible when the workflow changes — a new branch means editing nodes.
Email TriggerExtractRouteScoreEnrichCreate recordVCbrain

Tap a node for details

VCbrain-native agent

Adaptive · harness + MCP

A self-improving harness reasons over two grounds. Live tools arrive through the Model Context Protocol — each (Gmail, HubSpot, web) is an MCP server the agent calls on demand. And your own context: the files you authorise are distilled, locally, into a company knowledge graph that never leaves your infrastructure. Every IC correction feeds back to sharpen the next run.

Best for.
Funds without n8n; teams that want one product, not a stack.
Tradeoff.
Less predictable — needs robust logging and monitoring on our side.
VCbrain agent · your infraGmailHubSpotWebMCPMCP clientAuthorised filesstays on your infraDistilCompany knowledge graphHarnessPlan · Act · Observe · ImproveStructured recordIC memoIC corrections → learns

Tap a node for details

Lead with n8n for funds already on it. The native agent is the default for everyone else — and the long-term moat.

07Business model

Implementation-led SaaS.

Land on a fixed-price setup project that earns trust and reveals each fund's needs. The recurring SaaS fee is the business. After ~5 funds the onboarding is productised — pure recurring margin.

Phase 1 — Setup project

€5–10K

Fixed-price project with defined deliverables: deal flow automated, CRM connected, precheck running. Typically 4–6 weeks. Earns trust and reveals the fund's specific needs.

Phase 2 — Maintenance SaaS

€500–1K / mo

The ongoing system runs, updates and improves. After 5 funds this is pure recurring margin — no new consulting work per customer. The business scales here.

Phase 3 — Intelligence upsell

€1.5–2K / mo

Source quality scoring, founder hotspots, the pattern library. Sold after 60+ days live, when the data exists to make the case. Higher price, higher defensibility.

Unit economics at 10 funds: €70–100K setup + €60–120K ARR maintenance + €180–240K ARR intelligence = €240–360K ARR before any new customer acquisition.

08Market

Beachhead to every company in the world.

Beachhead

11,000+

VC funds globally

€500 / mo per fund

€66M ARR at 10%

Tech-forward funds on n8n. High pain, high willingness to pay, fast to close.

Expansion

500,000+

PE, family offices, angels

€200 / mo per fund

€1B+ ARR

Same problem, different asset class. Pattern library and deal memory translate directly.

Endgame

Every company

that needs AI automation

€100+ / mo per fund

€100B+ market

Company brain as infrastructure primitive. Every business in the world.

09Traction

Built in 24 hours. Deploying with a live design partner.

27,588

entities ingested

100

conflicts detected

+51%

harness improvement

1

live design partner

  1. Apr 2026Qontext HackathonBuilt v0.1 — context base, conflict detection, self-improving harness, full UI. 24 hours solo.
  2. Apr 2026Ostwerk design partnerReal VC fund. Real deal flow. Live deployment of the ingestion pipeline via existing n8n setup.
  3. May 2026EnterpriseBench validation130MB enterprise dataset remapped to the VC domain. 27,588 entities, harness iter_0 → iter_7 in one session.
  4. May 2026Interactive showcase shippedPublic walkthrough of the company brain — deal-flow precheck, chat-your-fund, source intelligence, and a live interview → research → brief pipeline.
  5. May 2026n8n community nodeOne-click integration into any existing VC workflow. No code changes. Target: 10 fund pilots.
  6. Jun 2026Productionise + scale the teamShip the n8n workflow SOP, an agent-native harness-structure MVP, and stand up our FDE (forward-deployed engineering) team.
10Team

Built by operators who hit the problem themselves.

VCbrain started as a tool Ostwerk built to solve its own deal-flow problem. The first customer comes from a known DACH peer fund — not cold outreach.

CZ

Claire Zhu

Product & GTM

Drives product vision, the sales sequence, and the first DACH fund relationships.

JW

Jiantong Wei

Engineering & Architecture

Owns the technical architecture — ingestion, the fact graph, agents and the security model.

See your fund's brain, live.

A 30-minute walkthrough: deal flow automated end to end, a credible EU data-security architecture, and a scoped setup proposal. Built for 5–10 person DACH funds doing 100+ deals a year.

We reply within one business day. No deck required.

VCbrain — The company brain for VC funds.