FundExecs OS

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.

What it replaces	What it becomes
DealCloud / Dynamo	Native deal pipeline + graph
Carta / Juniper Square	Fund admin + waterfall engine
Affinity / HubSpot	Relationship intelligence graph
Hebbia / AlphaSense	AI-native document analysis
Notion + Slack + Zapier	Unified workflow + agent layer
Visible / Passthrough	LP reporting + capital events

---

Project Status

🔴 Pre-Alpha — Seeking Contributors

What exists today:

• ✅ Full database schema (PostgreSQL / Supabase)
• ✅ API contract layer (native REST)
• ✅ WebSocket event stream architecture
• ✅ Agent definitions and capability specs
• ✅ UI component library spec
• ✅ Avatar animation protocol
• ✅ DevOps observability spec

What landed in the first build (2026-06-18):

• ✅ Next.js + TypeScript + Tailwind repo scaffold (single app: app/ + lib/)
• ✅ Full Postgres/Supabase schema as 11 versioned migrations, RLS on every table
• ✅ Org-membership tenancy model + six-agent seed catalog
• ✅ Typed data layer and hub/agent/event catalogs in lib/

What is being built next:

• 🔧 Deploy schema to a live Supabase project
• 🔧 Backend API services — the /prompt → /task → internal handoff → /approve loop
• 🔧 AI agent engine (Analyst, Associate, IR, Portfolio Ops, Diligence, Fund Admin)
• 🔧 WebSocket event gateway (Realtime over task_events)
• 🔧 Animated 3D workspace (Three.js + GSAP)

---

Architecture Overview

FundExecs OS is structured around four operational hubs, six AI agents, and three native data graphs.

The Four Hubs

Hub	Purpose	Key Modules
Build	Define identity and foundation	Profile · Thesis · Brand · Entity · Track Record · Team · Documents
Source	Manage pipelines and relationships	LP Pipeline · Debt & Hybrid · Partners · Providers · Deal Pipeline
Run	Evaluate and manage active deals	Strategy · Diligence · Underwriting · Stress Test · Risk · Outreach · Campaigns · Evaluate
Execute	Operate assets post-closing	Closing · Capital Events · Asset Management · Reporting · Exit
Network	Relationship-capital directory	Standalone side-rail page (/network)
Search · Marketplace · Meetings	Cross-cutting destinations	Standalone side-rail pages (/search, /marketplace, /meetings)

The Six AI Agents

Agent	Role
Analyst	Ingests deal data, financials, market comps — produces pro formas and valuations
Associate	Coordinates workflows and task execution across all hubs
Investor Relations	Manages LP communications, capital calls, and reporting
Portfolio Ops	Monitors asset KPIs, budgets, capex, and variance alerts
Diligence	Parses documents, flags risks, produces diligence memos
Fund Admin	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.

---

Tech Stack

Layer	Technology
Frontend	Next.js · React · Tailwind CSS · Three.js · GSAP
Backend	Node.js · Python · Native REST · Event-driven task engine
Database	PostgreSQL · Supabase · Redis
Storage	S3 for documents
Infrastructure	Vercel · Cloudflare · AWS · GitHub Actions
Observability	Prometheus · Grafana · OpenTelemetry · Sentry
Security	JWT · Row-level security · Encryption at rest · Audit logging

---

Database Schema

The core schema covers the full private-market lifecycle:

-- Core entities (excerpt)
principals · organizations · investors · deals
assets · capital_events · relationships · ai_agents · tasks · marketplace

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.

---

API Contract

All endpoints are native REST. No external SDKs.

Endpoint	Method	Description
/api/prompt	POST	Accepts user prompt; routes to hub and agent
/api/task	GET	Lists workflows and status
Internal handoff packet	—	Transfers structured context between agents
/api/approve	POST	Captures user approval; triggers automation
/api/report	GET	Retrieves task output and analytics
/api/agents	GET	Lists active agents and workloads
/api/graph/relationship	GET	Returns relationship graph
/api/graph/deal	GET	Returns deal graph
/api/graph/capital	GET	Returns capital graph

Full API spec available in /docs/api-contract.md

---

User Flow

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.

{
  "event": "task.progress",
  "agent": "Analyst",
  "hub": "Run",
  "task_id": "uuid",
  "progress": 0.65,
  "message": "Underwriting model updated"
}

Key event types: task.created · task.progress · task.completed · task.handoff · approval.requested · approval.response · graph.update

---

Animated Workspace

Agents are visualized as avatars in a 3D spatial workspace built with Three.js + GSAP.

Agent	Color	Motion Style
Analyst	Cyan	Precise, analytical
Associate	Indigo	Coordinated, rhythmic
Investor Relations	Gold	Smooth, communicative
Portfolio Ops	Green	Grounded, operational
Diligence	Red	Sharp, investigative
Fund Admin	Silver	Structured, methodical

Each avatar responds to WebSocket events in real time — spawning, executing, handing off, and awaiting approval visually.

---

Local Development

Prerequisites: Node 20+, npm, and the Supabase CLI.

npm install                 # install dependencies
cp .env.example .env.local  # fill in Supabase URL + keys
npm run db:start            # local Postgres + Auth + Realtime; applies migrations + seed
npm run dev                 # http://localhost:3000

npm run typecheck and npm run lint keep the tree healthy. See CONTRIBUTING.md for the build-order discipline and migration workflow.

---

Roadmap

• Frontend scaffold (Next.js + Tailwind)
• Database schema + RLS (versioned migrations)
• Auth + organization onboarding
• Backend API (prompt → plan → approve → automate loop)
• WebSocket event gateway (Realtime over task_events)
• Real AI agent engine — Claude-powered multi-step plans (claude-opus-4-8)
• AI Agent Copilot — primary surface (prompt → plan → step cards → approve & automate)
• Command Center dashboard — organizes the Copilot's output
• Product visual system (warm-black/gold, Space Grotesk / DM Sans / JetBrains Mono)
• Build Hub — Profile module
• Surface step deliverables as first-class artifacts (IC memos, models, risk reports)
• 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.

We are actively looking for:

• Full-stack engineers (Next.js / Node / Python)
• AI/ML engineers (agent orchestration, LLM integration)
• Private-market operators willing to test and give feedback
• Designers with experience in data-dense interfaces

To contribute:

1. Fork the repository
2. Check open issues and the roadmap
3. Open a discussion before starting large features
4. Submit a pull request with a clear description

Please read CONTRIBUTING.md before submitting.

---

Philosophy

Every tool in private markets was built to solve one problem. FundExecs OS is built to solve the system.

This platform is designed with three commitments:

1. No external dependencies for core intelligence — AI agents, graphs, and workflows run natively inside the platform.
2. Operators first — every design decision is made through the lens of someone who has actually run a deal or managed an LP relationship.
3. Transparency — built in public, versioned openly, governed by a clear contribution model.

---

License

MIT License — see LICENSE for details.

---

Contact

Built by a private-market operator. Questions, partnerships, or early access: [your contact / email / Twitter]

---

FundExecs OS is pre-alpha. The architecture is designed. The build has begun. If you see what this becomes — come build it.