This section explains how an AI agent uses your Klav vault: with its own identity and only the permissions you grant, asking you when the policy says so, and using secrets without seeing them. If you are an agent reading this, How agents work says exactly what you can and cannot get.
- How agents work: identity, the six words,
allow,askanddeny, and the permissions. - Connect an agent:
klav client connect,klav mcpand the skill. - Approvals and tasks: when an agent waits for you, and one approval for a whole job.
- Use secrets without seeing them:
klav agent run,injectandfill,klav.tomlandklav run. - The activity log: what each agent did, and what it saw.
- Pause, change or revoke an agent: stop every agent at once, change an agent’s permissions, or take its access away.