Your AI employees do the work.
Figs shows you what they did.
The AI employees your company puts to work — an accountant, a recruiter, a support rep — report what they own, what they've done, and what needs a human, in one place your whole team can see.
01The problem
You can't watch
30 consoles.
Companies are putting AI agents to work as back-office employees — reconciling invoices, screening candidates, closing the books. But the work happens in terminals nobody's watching. When an agent needs a human, it waits. When it finishes, there's no record.
Work hidden in consoles
Each agent runs in its own terminal. No one place shows what they're all doing right now.
Handoffs that silently wait
When an agent needs a human to decide, it stalls in a window nobody has open.
No shared record
When the work is done, there's nothing the rest of the team can see, trust, or audit.
02Who it's for
For AI employees with a real job.
Figs is for the agents your company actually employs — an accountant that closes the books, a recruiter that screens candidates, a support rep that clears the queue. Purpose-built agents that own a function — not a personal coding copilot.
Accountant
Reconciles the books, flags anomalies, closes the month.
Recruiter
Screens candidates, schedules interviews, drafts offers.
Payroll & HR
Runs payroll, applies changes, answers comp questions.
Customer support
Clears the queue and escalates the calls only a human should make.
Sales & RevOps
Enriches leads, keeps the CRM clean, chases renewals.
Data analyst
Pulls the numbers, builds the dashboard, answers the question.
If an agent owns real work and reports to a human, it belongs on the chart. Build those agents with OpenFigs.
03What it does
Your whole AI workforce on one chart
Every agent you run, grouped by the department it works in, with the human who owns it — a live org chart, computed from what your agents push. No wiring, no diagramming.
Know the moment an agent needs you
When an agent hits something only a human can decide — an approval, a question, a blocker — it stops and calls for you. Every call lands in one inbox your whole team can see, so nothing waits silently in a terminal nobody's watching.
Every run, on the record
Each agent reports what it did — runs, outputs, and the artifacts it produced, rendered inline so you read the real report, not a log line. An append-only history your whole team can audit; Figs never writes back, so the agent stays the source of truth.
Built on an open protocol. Works with any agent
Agents publish through the open `.figs` protocol with one `figs push`. Claude Code, Codex, Cursor — if it can write a folder, it shows up here. Build → report → govern, all open.
04The stack
Build → report → govern.
Figs is one product, built on an open standard. Build your agents with OpenFigs, let them report through the open .figs protocol, and govern the whole workforce in the app. The protocol is open, so nothing locks you in.
The open skeleton for production AI employees — structure, memory, and reporting built in. Stand up an agent that does the job and reports to Figs out of the box.
Explore OpenFigsAn open, MIT-licensed protocol your agents publish state through. One figs push and what they own, did, and need is live.
Read the specThe org chart, the calls inbox, and the activity record your whole team sees. The piece you're looking at right now.
Open the app05Trust
Safe to point at the back office.
Figs sits next to agents doing real, consequential work — so it's built to earn that trust. It watches, it doesn't touch.
Read-only by design
Figs mirrors what your agents push and never writes back. Your repos, tools, and data stay untouched.
You decide what's shared
What an agent reports is between you and that agent. Figs only ever sees what it chooses to push — nothing more.
Not another AI
No model, no agent runtime, no access to your code. Figs is the layer on top of the agents you already run — Claude Code, Codex, Cursor — not a replacement for them.
Open protocol, no lock-in
Your history lives in .figs — an open, MIT-licensed format on your own machine. Read it, export it, or walk anytime. The protocol and CLI are open source; the app is the hosted layer on top.
06Get started
See your AI workforce
in under two minutes.
Figs goes on top of the agents you already run.
Bring your agents
Already running Claude Code, Codex, or Cursor? You're ready. Starting fresh? OpenFigs scaffolds an AI employee in minutes.
Create your account
Click the button below — free, with unlimited agents for your whole team.
Connect your agent
Spin up a workspace and hit copy. Paste the setup prompt into your agent — it registers itself and pushes. No CLI to wire up.
Watch them work
Agents land on the org chart, activity streams in, and anything needing a human shows up in your calls inbox — visible to the whole team.
07Pricing
Free. Unlimited
agents. Forever.
Run as many agents as you want — every figs pushis free, and you're never billed per agent. Bigger teams get paid plans later — always per person, never per agent.
- Unlimited agents — every push, free forever
- Org chart, calls inbox, and activity record
- Free to start on the hosted app
- Your data in .figs — an open format, no lock-in
FAQ
Questions & answers.
Any agent that can write a .figs/ folder and run figs push — Claude Code, Codex, Cursor, and others. The protocol is open and MIT-licensed, so support isn't tied to any one vendor.
Figs is a hosted app at app.figs.so. Your data isn't locked in, though: each agent's record lives locally in an open .figs/ folder (the protocol and CLI are MIT), so you can read or export it anytime.
No. Figs is a read-only window — your agents push their state and Figs mirrors and displays it. It never writes back. Your repo and your agent stay the source of truth.
Each agent keeps a local .figs/ folder as its durable record and pushes to your workspace. That open, local-first file is the source of truth — yours to keep, read, or export anytime.
Agents are great at doing the work but bad at being legible to a team. Figs gives your whole org one place to see what every agent owns, what it has done, and when it needs a human.
Agents are free and unlimited, forever — you're never billed per agent. The protocol and CLI are open source (MIT). The hosted app is free to start; bigger teams get paid plans later, priced per person.