Total recall.
Twenty years of deals, passes and people — queryable in seconds. People leave. The fund’s memory doesn’t.
Ambient workflows.
Meeting prep, deal briefs, LP answers — finished before you asked. Running in the background, every day.
AI your LPs can sign off on.
Permissions to the row. PII masked. Every query logged.
This is already happening
Your team is already vibe-coding with your fund’s data.
An associate pastes a cap table into ChatGPT before an IC. A partner has Claude draft LP paragraphs from a board pack. Someone builds a beautiful portfolio dashboard in Claude Code — an HTML file that’s stale by Friday and lives on one laptop.
The work is real. The setup isn’t:
Same question, three different numbers.
Vibe-coded solutions that die on arrival — nobody can host, update or share them.
The answers are only as good as what one person remembers.
portfolio-dashboard.html
⚠ Stale · updated Fri
Q2 portfolio overview
last synced 6 days ago · on 1 laptop
The product
One layer between your systems of record and every AI tool your team uses.
CRM · Drive · Notion · email · messaging → VICUNEA FUND GRAPH → Claude · ChatGPT · Cursor · your agents
Below it: your sources, untouched and still the systems of record. Above it: whatever AI tools your team already prefers. In the middle: every company, person, deal, document and conversation your fund has ever touched — extracted, entity-resolved and connected, kept current automatically, and served through one governed MCP.
The result: deterministic, auditable AI on fund data — instead of eight people improvising.
Why partners care
Everything your fund ever learned, working in every meeting.
Walk in impossible to out-prepare.
Before a founder call: every email, meeting note and touchpoint your fund has ever had with them — one question away.
Source — CRM, email, notes
“Haven’t we seen this space before?”
You had. Twice. The graph pulls the memos and why you passed — during the IC, not after.
Source — Drive, Notion, CRM
Board packs you can interrogate.
“Which portcos slipped on plan this quarter?” Answered across the whole portfolio, with the page it came from.
Source — board packs, reporting
Portfolio asks, matched to your network.
A portco needs a CFO, a customer, a co-investor. The graph finds who your fund already knows.
Source — CRM, email, notes
The DDQ, drafted by lunch.
Every number traced to a document. LP coverage gaps visible before LPs feel them.
Source — fund docs, CRM, email
Apps that stay alive.
The dashboards your team vibe-codes — hosted on the graph, always current, access-controlled.
Source — everything above
Where the graph leads
A competitive deal. Lost. Three weeks later: a partner had sat on a board with the company’s angel. For six years. The connection existed. Nobody could see it.
The Fund Graph already resolves every person your fund has ever touched. Next: the map of what that adds up to.
→
Who your fund knows — and how warmly.
→
The warm path to any founder, fund or LP. A few hops, not a cold email.
→
A target’s cap table, mapped against your network before the first call.
The Relationship Graph. Same foundation. In build with design partners. Live Fund Graph → first in line.
Join the early-adopters list
Getting started
Couple of steps, raw sources to running fund.
From “we have Attio, a Drive and twenty years of history” to “the whole team ships on it” — inside two weeks.
01
Connect your sources.
A forward-deployed engineer maps your CRM, Drive, Notion, email and notes. We handle ingestion.
02
Vicunea builds the graph.
Entities resolved, relationships mapped, every fact linked to source. Kept current as your data drifts.
03
Define access.
Role-based controls for people and agents. PII masked. Preview as any role before you ship.
04
Switch on workflows.
Meeting prep, deal briefs, portfolio updates — from the fund library, or built for you. Audit baked in.
05
Ask from your tools.
One MCP into Claude, ChatGPT and Cursor. Per-user permissions. Full audit trail.
One product. One contract.
Book a working session
$5,000 one time build
+
$1,000 / month maintenance subscription
Platform fee plus usage. One contract.
Every plan includes
Forward-deployed engineer onboarding — live in 14 days
The Fund Graph: ingestion, entity resolution, graph — kept current
One governed MCP into Claude, ChatGPT, Cursor
Role-based access to the row · PII sandbox · full audit trail
The fund workflow library (meeting prep, deal briefs, portfolio updates)
Hosting for the apps and workflows your team builds
Usage, transparent
LLM — bring your own model, or pass-through at cost
Storage & compute — metered at typical fund volumes
Pricing
Priced to your fund. Line items included.
Tell us two things — how big the fund is, how much it touches. We show the plan, what’s in it, and what the same shape costs to build yourself.
FAQ
What do I get in 5 minutes / 24 hours / 7 days / 14 days?
5 min — the demo, on a real (anonymized) fund graph. Ask it anything, watch the citations come back. 24 h — your first source connected read-only; first cited answers from your own data in Claude. 7 d — core sources ingested, entities resolved, roles defined. First workflows on. 14 d — the whole team live. One MCP, per-user permissions, audit running, workflow library on.
How does onboarding work?
A forward-deployed engineer runs it, end to end. Week 1: maps and connects your sources, builds the Fund Graph, sets access with you. Week 2: switches on workflows, plugs the team’s AI tools into the MCP, hands over. You attend two working sessions. That’s your total lift. Running in 14 days — that’s the contract, not the aspiration.
Can’t I do this on my own?
You can start. Here’s what you’re signing up for: Connectors break monthly — re-auth, API changes, schema drift. Someone owns that forever. Entity resolution — matching “J. Novak” across five systems — is the hardest problem in fund data. Prompting doesn’t solve it; engineering does. Per-query costs compound — without an ingest-once layer, every question re-reads your files. Slow, expensive, unverifiable at scale. Governance can’t be bolted on — RBAC, PII masking and audit trails have to be in the layer from day one, or your AI story fails its first LP question. Build your edge in-house. Don’t build plumbing.
I’ve already built custom agent workflows. Will they be lost?
No — they get better. Your agents currently improvise: scraping files, guessing at duplicates, breaking when a schema moves. Point them at the Fund Graph instead and they run on infrastructure built for agents by design: resolved entities, cited facts, stable interfaces, permissions enforced underneath. Same workflows. Better answers. Nothing to rebuild.