Topics

Use these guides to understand how Agents run, control what they can access, and decide which changes require a new deployment. If you are new to OpenClaw Enterprise, start with Concepts or set up an OpenShell Codex Agent.

Agent

• Agent overview: identity, supported operations, and what happens when you deploy or stop an Agent.
• Presets: reuse launch settings and fill variables when creating an Agent.
• Compute: where Agents run and which execution modes each bundled Driver supports.
• Harness: how OpenClaw or Codex runs the model and authenticates to it.
• Remote Harness: how dedicated execution separates Gateway and Harness storage, and how documents, Memory, Skills, and attachments cross that boundary.
• Workspace files: edit an active Agent's instruction and identity files.
• Runtime logs: read Pod status, restarts, Events and redacted container output for a deployed version.
• Console walkthrough: understand each Agent detail control, status, and editing surface.
• Troubleshoot: follow a failed deployment, check model access, or recover after an uncertain request.

Security

• IAM: who can sign in and which resources they can use. For sign-in methods and sessions, see authentication.
• Namespaces: how Agents and configuration belong to teams or tenants.
• Sandbox: what isolates a workload and where the current limits are.
• Secrets: how to use credentials without putting their values in configuration.
• Audit Log: which security and management actions are recorded and who can access those records.

Capabilities and configuration

• Plugins: which curated tools an Agent may use and how approval rules work. Follow Configure to change an Agent's selections.
• Configuration: what belongs to an installation, a Namespace, or an Agent draft.
• Agent revisions: change a draft, deploy it, and inspect the snapshot the Agent runs.

For setup instructions for a specific vendor or Driver, use Integrations. For installation-wide work, use Deploy & Operate.