Agents and readiness
AI OS gives each local agent a product surface without pretending that all agents share one runtime or capability contract.
Hermes
Section titled “Hermes”Hermes owns the local conversational and operator workflow exposed by its agent page and Intelligence portal. The UI reports session, model, memory, persona, skill, bridge, and recovery state from bounded local contracts.
Its shipped workflows include Chat, Mission Control, Documents, Pantheon, Ministry, model intelligence, Voice, and the Intelligence portal. Read the Hermes guide before treating one subsystem’s readiness as proof for another.
Claude Code
Section titled “Claude Code”Claude Code remains a distinct local agent surface. Its execution cockpit can run bounded workspace, shell, Git, process, and file operations when local readiness passes. It does not merge Claude Code state into Hermes sessions.
OpenClaw
Section titled “OpenClaw”OpenClaw has its own readiness and local-action boundary. Deploy, spawn, and action controls execute through the configured local OpenClaw dependency when available. Its state should be interpreted from the OpenClaw surface rather than inferred from another agent.
Readiness is specific
Section titled “Readiness is specific”A visible agent card does not prove execution readiness. Check the current status, selected model, available local bridge, and recovery message on the agent’s page. Public-demo agent states are illustrative only.
See Claude Code and OpenClaw for their execution boundaries. Dream and Trend Finder use the separate OpenAI subscription runtime when that provider is selected; Hermes chat does not.