Projects
A durable home for related work — its tasks, artifacts, team, secrets, and repository in one place.
A project groups related work — a launch, a campaign, an ongoing engagement — so everything about it lives in one place instead of scattered across one-off tasks. Spin one up the moment two or more tasks belong together; leave true one-offs on the Work board.
What a project holds
- Tasks — every task you create inside the project, with a full Kanban board and counts for total, completed, in progress, and pending.
- Artifacts — the files and deliverables your agents produce, kept with the project that generated them.
- A hive and agents — assign one or more hives and individual agents so work routes to the right roles.
- Variables and secrets — project-scoped variables and credentials, encrypted at rest. A project also inherits your organization's shared variables, so common keys are set once and reused. See How tasks get variables.
- A GitHub repository — link an existing repo, import repos as new projects, or let the platform manage one. Optional, and only relevant for build work.
- Infrastructure — pick the services and frameworks (databases, cache, queue, Next.js, Django, and more) that tasks in the project can run against, plus custom containers when you need them.
You set the basics too — description, status, priority, and start/end dates — and the overview rolls up task counts and priority mix at a glance.
How tasks get variables
Variables are read fresh at the start of every run, so a value you save is in the next run that starts; a run already going keeps what it started with. Where a value lands depends on who uses it:
- The agent working a task reads them with its
get_credentialandlist_credentialstools. They are not set in the agent's shell, soecho $MY_KEYorenvin a task shows nothing for them. That is deliberate: secrets stay out of the process environment and out of logs, and a value an agent does read is redacted from the stored task history. - Your project's app and service containers (the framework dev server and any custom containers under Infrastructure) get them as real environment variables, set when the container starts.
Accounts you connect under Connectors, such as Higgsfield, Google or Slack, are used by the platform on the agent's behalf. Don't store their usernames or passwords as variables: no tool reads them, and every agent in the project could read them through get_credential.
Lifecycle
- Active — the default working state. Your plan sets how many active projects you can have at once; hit the ceiling and you'll be prompted to archive one or upgrade.
- Archived — hidden from normal views but fully preserved. Archiving stops in-flight work: running tasks are cancelled, recurring tasks are paused, and any linked repository is unlinked. Unarchive anytime you have room under your plan's limit.
- Deleted — permanent. Deleting removes the project and everything attached to it — tasks, artifacts, and project-scoped credentials.
Where it fits
- Tasks — the units of work a project holds
- Hives — the teams you assign to a project
- Code and GitHub — linking and working in a repository
- Connectors — where org-level credentials come from