Projects
A project is what Jaah works on: one or more repositories, such as an API and its web app, or no repository at all. Each repository has its own integration branch, and one of them is the primary: the gate, the stack and the ports run there. A project's settings say where its worktrees run, how runners check and deliver their work, and who may use it. These guides cover the jobs you do on a project most often. The ideas behind them are in How projects work.
- Add a project — connect its repositories and give it worktrees to run in.
- Give someone access to a project — grant a person or a group a role on one project.
- Add an environment variable — give a project's tasks a value, typed in or read from a credential.
- Change a project's integration branch — point a repository at another branch, so task pull requests open against it.
- Change or delete an environment variable — change a variable's value or remove it from new slots.
- Publish a Live Cast — publish a project's tasks to a public page, with rules and per-item overrides.
- Use the merge lane — let a lane test and merge the tasks' pull requests together, and what a failed gate does.
- Release to production — move merged changes from Dev to Production with a release, approvals, schedules and freezes.
- Retry a failed release — send a release that failed before its merge again, on the same pull request.
- Approve or reject a release — read what approvers are asked to check, then let a release through or turn it down.
- Release with a change ticket — type the Jira or ServiceNow ticket a release step asks for.
- Connect your own domain — show an Easy Mode app on a domain you own, with the DNS records to add.
- Choose who background jobs run as — pick the member whose Git access Jaah's own work pushes with, for the org or one project.
- Set up a project's pipeline — add, rename, move and remove stages, set who may do each step, approvals, by hand or by itself, and freezes.
- Start a pipeline from a preset or another project — pick a ready-made pipeline at create, or copy another project's, then review and save.
- Share a pipeline with a template — keep one pipeline for many projects: lock what they can't change, edit a draft, review and publish.
- Link a project to a template — give a project a template's stages, change only its open fields, unlink to edit everything.
- Update projects a publish left behind — find linked projects a publish skipped, fix why, and retry with Update again.
- Set your org's pipeline rules — check every project's step into production on save: approvals, never by itself, pressed by maintainers, as a warning or a block.
- Fix a pipeline an org rule blocks — read the stage card's message, change the step into production and save again.
- Choose a basic pipeline — pick PR into integration branch or Integration + Release, and set the release branch and addresses.
- Add a gate to a pipeline step — make a merge or release step run checks first, and verify a stage after a change arrives.
- Set a stage's deploy target — tell Jaah how a stage goes live: your forge, Argo CD, a webhook or a script.
- Ship only some changes in a release — tick the tasks that should go out and leave the rest waiting.
- Release a task that changed no code — ship a task finished without a pull request in the next release, like merged code.
- Require approval before a step moves on — set how many approvals a step needs, who approves, a timeout and instructions.
- Limit when a step runs — hold a step with a wait after arrival, or to chosen days and hours.
- Release changes that have not been tested — ship what waits for the merge lane's tests now.
- See which changes are still in testing: spot the flask on changes the merge lane is still testing, and what a release ships of them.
- Schedule a deploy — pick a date and time in When so a release starts later by itself.
- Replace a project's Jaah repository — put a project Jaah manages on your own repository, moving its code and history.
- Give a project's agents an MCP server — link an MCP connection so the project's sessions can use its tools.
- Review or purge a project's memory — read what runners learned, retire a unit, or purge one that holds sensitive data.
- Turn project memory on or off — choose whether a project's tasks draw on what earlier tasks learned.
- Set how runners push to a repository — read only, branch prefix, force push, and the test gates their work must pass.
- Squash task pull requests into one commit — turn on Squash task PRs so each task lands as one commit named after it.
- Compare agent setups with benchmark mode — re-run a sample of merged coder tasks on another setup and read who won.
- Delete a project — check what is still running, then remove the project with its tasks and slots.
- Hide a project from your board — take a project off your sidebar and board, for you only, and show it again.
- Rename a project — change a project's name in place, in the trail at the top of its page.
- Tell runners about a project — standing agent instructions, the tech stack and notes every task can read.
- Add skills to a project's agents — keep skills in the project's repository so its agents load them; what to use where a skill can't go.
- Restore an earlier version of a pipeline — bring the stages, steps and freezes back to an earlier save, from the custom pipeline's History.
- Set a project's routing mode — let routing suggest or pick each task's runner, cap its hours, and turn on blind review.
- Add or remove a project's repository — give a project a second repository, pick the primary, and remove one it no longer needs.
- Use a repository on your own GitLab server — connect a self-managed GitLab, then add its repository with GitLab as the forge.
- See what a project changed from its defaults — read the On marks in the Settings sections, and put a setting back.
- Set a project's default persona — choose the persona a project's new tasks start as, and its default model.
- Let a project's sessions reach another site — add an internet name a project's MicroVMs may reach, or block a default one.
- Change how big a project's session VM is — give a project's MicroVMs more vCPUs, RAM or disk, or inherit them.
- Run a project's outcome scan now — read when merged work was last scanned for reverts and fixes, set how often, and start a scan.
- Set how often your projects are scanned — your organization's Outcome scan interval, for every project without its own.
- Find a setting in a project's settings — type a setting's name in Search settings and jump to it, outlined in its section.
- See which projects need you — sort the Projects list by questions, running or queued tasks, and go to the project that's waiting.
- Finish a new project when no node or agent is ready — add a node from New project without losing your answers, or create the project before you have a model account.
- Fix a project's push warning — read Tasks can't push here. and fix it with Fix.
- Check a hosted app's usage this month — read an Easy Mode app's requests, egress and database size against its limits.
- Retry a hosted app's failed setup — run an Easy Mode app's setup again with Retry when it didn't finish.
- Fix a variable that does not resolve — read why a variable marked Not resolved gets no value and point it at a secret that works.
- Merge tasks through a freeze — override a freeze holding the Merge step, with a reason recorded on the Audit page.
- Track a release until it deploys — connect Deploy tracking so a release stays Releasing until Production's deploy succeeds.
- Raise a project's budget when its tasks wait: read Spent this month on the Budget tab, then raise or remove the cap.
What every field on a project's tabs means is in the Projects reference. To give a project's tasks a credential, see Attach a secret to a project.