[ Documentation ]
The governed runtime for
enterprise AI agents.
Run the agents your teams already use. Govern every action they take. Claude Code, Codex, Cursor, OpenCode. Any model, frontier or open-weight. In your VPC, in xpander’s cloud, or fully air-gapped.
Start where you are
Every call an agent makes, to a model or to a company system, goes through xpander: validated, credentialed at the moment of use, budgeted, and tied to the person who asked for it. Risky actions wait for a named person’s sign-off. Pick the door that matches what you do.
User guide
You use Claude Code, Codex, Cursor or OpenCode
Connect your client to xpander once. A job you send to the background runs on xpander in an isolated session with your organization’s approved skills, and shows up as a task in Xpander Chat, where your team can follow it, approve held actions and continue the conversation.
Admin guide
You set the policy, or you operate it
Approved skills and models, budgets, approvals, credentials in the Vault, sign-in with your identity provider and the audit log live under Settings in Xpander Chat. Hybrid and air-gapped deployments install from a Helm chart into your own cluster.
Build with the API
You integrate xpander
One call starts a run on any agent and returns its result; tasks, threads and model usage are readable afterwards; skills, knowledge and workflows are managed over the same REST API. Python SDK, CLI, webhooks and MCP sit on top of it.