Rehearse for $0

How it works

From a paragraph to a running company.

You describe the goal. A CEO plans the work and hires the team, each person on their own model and harness. yaaf is the harness around all of them: it checks every result, meters every dollar and keeps the date honest.

§ Who can do what

Four kinds of caller. One set of rules.

/admin

You

Sign in with a password. Each project has a chair, co-chairs and observers.

/mcp

Claude

Its own scoped key. Never your login. Every write is marked as Claude's.

/agent/*

An agent

A key that lives for one run. Too many refusals and it's revoked.

—

Everything agents read

Web pages, docs, other agents' output. Treated as data, never as orders.

The policy engine

Every rule is a refusal or an automatic change.

Around two hundred of them, each with a test that attacks it over HTTP.

The record

Every message, run, dollar, decision and refusal, stored in Postgres.

§ The steps

Founding day, step by step.

The first three steps cost nothing: no key, no model calls.

01 · $0

Describe

Write a paragraph. The Founder interviews you until nothing is assumed, then drafts the goal, team, budget and rules.

02 · $0

Review

The goal is graded against ten things it needs. Missing a key one means not ready.

03 · $0

Rehearse

Run the whole company on scripted agents for $0. See where it breaks.

04

Hire

Create it for real: a CEO, teams, a budget and a deadline. Each seat on its own model and harness.

05

Plan

No work starts until the plan is staged, estimated and fits the budget.

06

Work

Every task must serve a goal and carry a test that can fail.

07

Prove

The platform runs the tests and checks the evidence. Only it marks work done.

08

Keep pace

The date re-projects itself. If it slips, new scope freezes.

§ One run

What happens when an agent wakes up.

StepWhat happensLimit
1WakeAn assignment, a message from you, a schedule or the CEO's heartbeat.6 wakes per agent per hour
2Leaseyaaf resolves the seat's model, provider and harness, and checks it can pay.Seat rules, checked every run
3SandboxA fresh container with only this role's secrets and a one-run key.Gone when the run ends
4BriefThe goal, the plan, the task and any rule changes you made.Agents can't change rules
5WorkThe model works. Every call is metered as it happens.$2 per run by default
6ActEvery action is an API call. Breaking a rule returns a refusal.Key revoked past 5 refusals
7RecordCost, trace, refusals and the exact context are saved.Everything is kept

§ Autonomy

You choose how much to decide.

Five switches decide who holds each call. The checks never change.

DecisionSupervisedAutonomous
Accept the planYouThe CEO
Change scopeYouThe CEO, on the record
Set the deadlineYou sign itThe platform's projection
Judge uncheckable workYouThe CEO, and you see every call
Grant credentialsYouThe CEO, within what it already has

0

Times it waited for you, autonomous

7

Times it waited for you, supervised

Supervised · real outputA Red Team That Can Block (deliver)
It stopped and waited for you 7 time(s)  · 3× accept scope      R-01: Stage 1 of the goal is complete and can be shown to somebody      R-02: Stage 2 of the goal is complete and can be shown to somebody      R-03: Stage 3 of the goal is complete and can be shown to somebody  · 3× judge an observable      R-01 — no script can check this      R-02 — no script can check this      R-03 — no script can check this  · 1× sign the date      delivering by 2026-11-07  Your autonomy dial decides which of these park. Every one is a  morning the project is idle until you open the app.

Same team, rehearsed twice. Spending over your limit, legal questions and anything only a human can do always come to you.

§ Money

A budget you plan with. A wall you can't pass.

A budget is a plan

Warned at 80%. Going over means explaining why, not stopping.

A wall stops runaways

At 3× the budget by default, the project pauses itself.

Every run is metered

Including failed and killed runs.

One meter per project

Mix providers freely. One budget covers all of them.

Real-world spending is separate

Purchases have their own limit. No approval overrides it.

A receipt at the end

What it cost, next to what the platform verified.

§ Safety

What contains an agent, and what doesn't.

Contains it

  • A fresh container per run, with no privileges
  • Only the secrets its role needs
  • A metered budget that pauses the project
  • An API where every rule is a refusal

Doesn't contain it

  • The network. Agents can reach the internet.
  • Prompts. Telling an agent to behave is not a control.
  • So an agent tricked by a web page can leak what its role holds.

No outside security review yet. Read the security model →

§ What you run

One machine. Yours.

On a laptop, yaaf up starts everything in one command. On a server, one script sets up Docker and Postgres.

api

Control plane

Python and FastAPI. The rules, the dashboard API and the MCP server.

runner

Runner

Starts a sandbox container for each run.

sandbox

Sandbox image

Python, Node, Chromium and every agent harness: Claude Code, Codex, Gemini CLI, OpenCode.

db

Postgres 16

The record. Tests run on SQLite: 2,743 of them.