Claude Code and OpenClaw

Claude Code and OpenClaw have separate AI OS pages, local executors, and readiness contracts. Neither is an alias for Hermes.

Claude Code

The Claude Code page combines Mission Control with a local execution cockpit. When local readiness passes, it supports bounded workspace, shell, Git, process, and file actions through reviewed operations.

Execution remains confined to the allowed workspace and command catalog. Results are capped and projected into visible success or recovery states. A route shell or visible button is not proof that a local dependency is ready.

OpenClaw

The OpenClaw page presents status, sessions, sub-agents, tasks, skills, and plugin metadata from the OpenClaw-owned local boundary. Deploy, spawn, and action controls execute through configured local OpenClaw action paths when their dependency checks pass.

If OpenClaw is missing, offline, or not configured for the requested action, the page should preserve that as a named recovery state instead of showing a false success.

Shared local safeguards

Both surfaces keep the local access posture while retaining request checks, same-run authorization, schema validation, timeouts, bounded output, safe errors, and path or operation confinement where required.

Normal bun run dev startup provides the local all-access posture. Legacy manual admin flags are compatibility details, not standard launch requirements.

Public demo

Public-demo agent cards are illustrative. They can show product structure and recovery language, but they cannot execute Claude Code or OpenClaw actions.

For conversational work, Mission Control, Ministry, Voice, and Intelligence, continue with Hermes, Voice, and Intelligence.