Product · Platform
covey
covey treats AI agents like employees: an identity of their own, an isolated workplace, time-limited access, a place in the org chart — and central oversight of all of them together.
How covey is built
One program, one database, isolated workplaces beside them. What is described here is not a statement of intent — this is how it runs.
The control plane
A Go program next to a Postgres database: planning, the register of agents, the task list, management of sign-in and credentials, the rule engine and the recording. It holds all the state and is always reachable.
The data plane
Isolated environments that are cleared away after the work; only the working directory stays. Deliberately kept simple and replaceable: if one is lost, it is rebuilt from configuration and working directory.
The protocol in between
Between the control plane and the workplace runs a protocol both sides can speak over. It is the point at which engines become replaceable — Claude Code today, Codex and educa alongside.
Your own servers
The data plane does not have to sit on the same machine. A built-in server is the default; further ones register themselves, the way you know it from GitLab.
Postgres as the anchor
State, tasks, permissions, queue, notifications and the agents' memory all live in a single database. No additional message service, no cache cluster, no second place where state is kept.
Swappable where it counts
Sign-in and credentials are replaceable: built in with our own means, or swapped for a company login and a central secret store. The built-in route is complete and not a stopgap.
For the whole organization
covey is not a tool for one person but a platform for several roles:
- IT & platform admins who run the workforce
- Team leads who own agents and prioritize tasks
- Security & compliance, responsible for rules and evidence
- Controlling that keeps an eye on cost and budgets
See covey in action
Sign in — or discover how the Companion app connects people and agents.