The AI-native operating system for private capital. Fifteen agents source capital, underwrite deals, manage LPs, and own the work across every hub — on a schedule, approval-gated by default.
An AI-native operating system for private-market participants — unifying relationships, deals, and capital into a single intelligence layer, with AI agents that execute workflows end-to-end.
4+ years running advisory for PE funds and family offices taught me one thing:
The work that kills you isn't the decisions — it's the information movement before them.
An AI-native operating system for private-market participants — unifying relationships, deals, and capital into a single intelligence layer, with AI agents that execute workflows end-to-end.
The Problem
4+ years running advisory for PE funds and family offices taught me one thing:
The work that kills you isn't the decisions — it's the information movement before them.
I spent 3 hours every day sourcing deals, chasing data across a dozen tools, only to find out the opportunity was smoke and mirrors. DealCloud for pipeline. Notion for notes. Excel for underwriting. Slack for follow-ups. Carta for cap tables. None of it talked to each other. None of it got smarter over time.
There has to be a better way.
FundExecs OS is that way.
What It Is
FundExecs OS is the system of record for private-market operators — combining multi-agent AI, real-time data graphs, and an animated workspace to replace 30+ point solutions.
It automates the 80% of time spent moving information so you can focus on the 20% that drives decisions.
Wallet and Earn Copilot Visual Refresh
The visual system now has an electric-blue enterprise console treatment on a
single bold, vibrant daylight palette (no day/night switch and no dark mode, so
wording stays legible on every device), compute-grid panels, a clear elevation
hierarchy, CSS-only activation streams, and an Earn copilot drawer that reads
like an active operating-agent console.
The animated SVG below is a lightweight demo placeholder for the README and
marketing surfaces. It shows the Pro selection flow and Earn routing feedback
without adding video binaries or capture tooling.
Visual implementation guardrails live in
docs/VISUAL_SYSTEM.md: blue accents for the
operating shell, neural blue for compute/copilot surfaces, status tokens for
semantic states, and centralized logo usage.
Handles waterfall calculations, fund accounting, and audit prep
The Three Graphs
Relationship Graph → Who knows whom; who invested or financed what
Deal Graph → Deals, targets, portfolio companies, SPVs, funds
Capital Graph → LPs, investors, lenders, family offices, banks
All graphs are native, first-party data structures — no external dependencies.
Full schema lives as versioned migrations in
/supabase/migrations — applied locally with
npm run db:start. Every table is org-scoped and protected by row-level
security.
User enters prompt
↓
AI parses intent → routes to hub + agent
↓
Task engine creates structured tasks
↓
Agents execute → animated avatars visualize progress
↓
Agents hand off work between each other
↓
AI recommends next move → requests user approval
↓
User approves / disapproves / regenerates
↓
Loop continues until new prompt
WebSocket Event Model
The workspace is live — every agent action triggers a visible event.
Persist deals/assets from workflows so the Command Center populates from real work
Capital Map — relationship temperature, thesis fit, warm-intro pathfinding + gated next actions (Affinity, rebuilt native)
Gate layer — Tier 1/2/3 control primitive so no action reaches a counterparty without sign-off
Unified Inbox — one AI-triaged stream for booking, messaging, and video; gated next actions, deep links into Command Center context, and a dashboard digest (mock-or-real channel adapters: Gmail, Slack, Calendly, Google Calendar, Zoom, Google Meet)
Three-graph data architecture (query layer + /graph/*)
Three.js avatar workspace
Remaining Source / Run / Execute hub modules
Contributing
This is an open build. If you've felt the same pain in private markets — or if you're a developer who wants to work on something with real-world complexity and domain depth — this is the place.
