Run a fleet of agent machines like a team.
Once many agents run in parallel, someone needs to own the machines, the spend, and the risky decisions. Workspaces put those controls in one place.
- Roles and ownership
- Owners, admins, and developers get different powers over docks, keys, and billing.
- Budgets and approvals
- Set spending limits and require a person to approve sensitive actions.
- Agents never approve charges
- Payment activation routes through human checkout. Service keys cannot reach billing or account closure.
- Keys for automation
- Automation uses dock_ prefixed service keys or scoped workload tokens instead of a person's session.
Questions
Which actions are dashboard only?
Payment, account closure, identity link management, organization membership, and webhook delivery inspection stay outside ordinary service key reach.
Use cases that fit
Coding agents
Run coding agents like Claude Code, Codex, and OpenCode in isolated, repository-aware Linux machines with Docker, previews, and cheap branches.
Learn moreBrowser and computer use agents
Keep Chrome profiles, logins, and session state on a persistent dock, with a live desktop for human takeover when the workflow needs judgment.
Learn moreQA and preview environments
Build branches, run services, expose private preview URLs, and keep screenshots, logs, and artifacts on a persistent dock per change.
Learn moreInternal agent platforms
Connect agents to internal tools and private systems with deny-by-default egress, brokered credentials, approvals, and a full audit trail.
Learn moreAgent evaluations
Branch one snapshot into many identical docks to run agents, prompts, or models side by side and compare results fairly.
Learn more
Tell us what your agents need to run.
Private beta is open. Describe your workload and we will reply about access.