Product
Kairo is an operating layer for real servers — natural language control, deployment intelligence, knowledge memory, and human approval built in.
Connect a host (SSH or Kairo agent), link GitHub when needed, then chat intent — clone path, branch, runtime, and services. Kairo plans, executes with tools, and streams operator-facing progress until the system is healthy and reachable.
GitHub and the operator client talk to the Kairo API. The AI decision engine coordinates knowledge retrieval, deployment execution, and monitoring against your infrastructure — with memory feeding future turns.
Configure deploy keys or HTTPS for public repos, analyze the workspace, install only after path confirmation, and complete only with verified remote evidence — Docker preferred, PM2 when appropriate.
Failure signatures and successful remediations live in the knowledge layer. High-confidence historical fixes can auto-apply without re-asking for the same error pattern.
Planner and executor select tools, respect mutability gates, and synthesize plain-language outcomes. Provider failover keeps the loop alive; billing quota is never confused with upstream LLM limits.
Push → webhook or chat goal → analysis → deploy → monitor → detect → RCA → confidence → recommend → approve → heal → complete. Every stage is visible; none are simulated as success without proof.
Yes (persist per server) or Read-only (prompt each time). No hard-block read-only errors — approval UI surfaces instead so operators stay in control.
GitHub OAuth and App webhooks, Docker, Linux hosts, major app stacks, and agent-based outbound connectivity for locked-down networks.
First-time chat setup is separate from webhook auto-deploy quotas. Profiles store path and restart commands so each push reuses the same operating path.