A software team in one Mac app. The decisions stay yours.
Great_cto runs a pipeline of specialist agents on your own Claude Code account. It plans, builds, reviews and tests the change, and stops where the decision is yours.
Nothing is deployed until you approve.ApproveSend back…
Illustration of the app, with sample data.
One feature, end to end, fully traced. The cost is model tokens. See the receipts →
Seven whole products built in the open benchmark, measured 10 July 2026. Reproduce it →
MIT. No seats, no account, telemetry off by default. You pay only your model provider.
The first two are historical measurements, not a benchmark of the current version.
Describe the task. Answer when it asks.
Describe the task
Press New task and say what you want: a new product, or a change in a project you already have.
The pipeline works
Each role does its part and hands over: the plan, the code, the review, the tests. You can watch every step on the task's timeline.
You decide
At a stop you see what was done and what was checked. Approve it, or send it back with a note.
Eight core roles. More when the task needs them.
A role is a responsibility, not a head count: a small fix does not call all eight. Tasks that touch money, personal data or migrations bring in their own reviewers.
| Role | Answers for |
|---|---|
| product-owner | the problem and what counts as done |
| architect | system design and trade-offs |
| pm | the plan and its dependencies |
| senior-dev | implementation |
| code-reviewer | independent code review |
| qa-engineer | verification and regressions |
| security-officer | risk and the ship gate |
| devops | the deploy you authorized |
You choose how often it asks.
Each project has one approval level.
| Approval level | It stops you at | Stops |
|---|---|---|
| ship-only | the deploy, with a brief of what gets built | 1 |
| product-only | what we build · whether it ships | 2 |
| gates-only default | what we build · the design · the deploy | 3 |
| strict | the design · code review · the deploy | 3 |
| auto | nothing in the pipeline | 0 |
Fintech, healthcare and government projects keep their security, compliance and ship gates at every level. A production release, a store submission or an over-the-air update always stops for you.
It cannot overspend quietly or approve itself.
A money cap on every run
Each run has a budget. When the cap is reached the run stops and asks: raise the budget, or stop.
A role signs only its own review
No role can write another role's verdict, and the app records who wrote each one.
Unknown is not a pass
A check that did not run shows as not verified. It is never drawn as green, as zero or as $0.00.
One project, four views.
Work
Tasks, what counts as done, and the timeline of each run.
Decisions
Everything that waits for you, in one list.
Knowledge
The project's documents and architecture decisions.
Insights
Task outcomes and what each project spent.
What it is not.
Not for a team
It is built for one builder. Several people working in one project is not on the roadmap.
Not a hosted app builder
It needs your own Claude Code account, and runs on your Mac only: there is no Windows or Linux version.
Not deterministic
Model output varies from run to run. The gates and checks exist because of that.
Not a certification
The compliance reviews are starting points for your own audit, not certificates.
Shipped lately.
4.1.0 · 17 November 2026
Start a task from the board, watch it work, approve its plan and cap what it spends, with no terminal.
4.0.0 · 14 November 2026
Great_cto now installs only what its maker signed, and checks it before it runs.
3.78.1 · 10 November 2026
Small fixes to the new Home and Cost pages.
Before you install.
What does it cost?
Great_cto is free, forever, and open source under MIT. You pay your model provider for the tokens your tasks use.
Where does my code go?
The board and project files stay on your Mac. Your coding agent sends prompts and the code it needs to its provider, under your account. There is no Great_cto account and telemetry is off by default.
Which coding agents does it use?
Claude Code runs the full pipeline. Codex support is experimental.
I installed it from npm. What now?
The npm package is retired. Install the Mac app to keep getting updates; your projects stay as they are.