AI agents for your whole backlog
Tickets in. Checked, merged work out.
pitcrew works through your project's tickets with AI agents. Each one plans or builds in its own sandbox, on your own computers or on pitcrew's servers in the cloud. Its work is checked, reviewed by a separate agent and merged when it passes, while you steer from the desktop app, the web app or the terminal.
Desktop app for Windows, macOS and Linux. The web app works on a phone too.
Who it is for
For people with more tickets than hours
Working on your own
Bring your repository and your own AI accounts. Agents work through your backlog in sandboxes on your computer or on pitcrew's servers, through the night if you let them, and stop to ask when only you can answer.
A team
One account for everyone, with teams, projects and roles: viewers watch, members file and steer, admins change the settings. Usage can be capped for a team or a project.
How it works
How a ticket gets done
Every ticket takes the same path, and you can see where each one is.
Say what you want
File a ticket, or describe something bigger as a request. Talk it through with an agent, agree on a plan, and it is broken down into tickets small enough to build one at a time.
Agents plan and build
Each ticket has a role. Architects and designers decide and design, developers and devops build, and a product manager keeps the backlog workable. Every run gets its own sandbox and its own copy of the code.
Checked, then reviewed
Your project's own checks run on the change. Then a separate reviewer agent scores the work, and only work that clears your bar goes on. The rest goes back with the reviewer's findings.
Merged
Approved work is merged into your base branch by a merge queue that checks it again beside everything else that landed. Prefer pull requests? pitcrew can open them on GitHub or GitLab instead.
When only you can answer, a ticket waits under Needs you and says what it is waiting for. You choose how much happens without asking: Hands-on, Balanced or Hands-off.
What comes with it
Everything around the agents
A prompt and a loop get an agent working. Keeping it working, day after day, takes the rest.
AI accounts
More than one AI account
Add several subscriptions or API keys to a project. When one reaches its usage limit, the ticket moves to another and carries on. Agents run on Claude Code, Codex or OpenCode.
Machines
Somewhere to run
Sandboxes on pitcrew's servers, or on your own Linux, Windows or Mac computer, as a virtual machine, a container or directly. A second computer can work the same board at the same time.
The board
A board that explains itself
A waiting ticket says why it waits, and a stopped one says what will release it. When a fix it needs lands, a blocked ticket goes back in the queue on its own.
Decisions
Decisions that stick
Architects write decision records, and every later run reads and follows them. You read and edit them in the apps.
Costs
Costs in view
Token use and cost by ticket and project, spending caps, the time each ticket spent in each stage, and how much got done with no one stepping in.
Your tools
Your tools, if you want them
Plan in Jira, deliver as pull requests or merge requests, and be told through a webhook or a command of your own. Nothing talks to an outside service until you turn it on.
The apps
Steer from wherever you are
Desktop app
Windows · macOS · LinuxThe board, requests, decision records and settings. It can also lend your computer to pitcrew: choose Use this computer, and your projects' sandboxes run on it.
Get the desktop appWeb app
any browser · phoneEverything the desktop app shows, in your browser. On a phone, answer an agent's question or see what merged overnight.
Open the web appTerminal
pitcrewOne command for all of it: pitcrew status for what is running, pitcrew why for why a ticket waits, pitcrew new to file one. Scripts can ask for JSON.
Why pitcrew
More than a personal setup
Plenty of people share how they run AI agents on their own. pitcrew is one whole workflow, with the parts a personal setup leaves out.
| What | A personal setup | pitcrew |
|---|---|---|
| Where agents run | In your terminal, while it stays open | In sandboxes on your computers or pitcrew's servers, whether you are there or not |
| A usage limit | Work stops until it resets | The ticket moves to another of your AI accounts |
| Checking the work | What you remember to ask for | Your project's checks, a separate reviewer, then a merge queue |
| Many tickets at once | One session at a time, started by hand | Several at once, in order of priority and of what each depends on |
| When something stops | You find out when you look | It says why and what will release it, and can tell you where you are |
| Working with others | Just you | Accounts, teams, roles and limits |
| Where you see it | Your terminal's scrollback | The desktop app, the web app on any screen, and the terminal |
Get started
Start with the apps
Get the apps
Download the desktop app, or open the web app in your browser.
Sign in
pitcrew is by invitation for now. Open the invite you were sent and choose how you sign in.
Add a project
Connect your code, choose where agents work, check the project and start. Then say what to build first.