Expand description
Agent runtime — orchestration, security, observability, cron, SOP, skills, hardware, and more.
Modules§
- agent
- approval
- Interactive approval workflow for supervised mode.
- browse
- Scoped one-level directory browser. Gateway (
api_browse.rs), CLI (src/browse.rs), and the future TUI directory picker all reach the same canonical implementation here. - channel_
targets - Build the configured channel targets section for system prompt injection.
- cli_
input - cost
- cron
- daemon
- doctor
- health
- heartbeat
- hooks
- i18n
- Fluent-based i18n for tool descriptions.
- identity
- Identity system supporting OpenClaw (markdown) and AIEOS (JSON) formats.
- integrations
- migration
- nodes
- observability
- onboard
- Onboard orchestrator.
- peers
- Peer-group runtime resolution.
- platform
- process_
stats - Self-process resource sampling — RSS (resident memory) and CPU%.
- rag
- RAG pipeline for hardware datasheet retrieval.
- routines
- Event-triggered automation (routines engine).
- security
- Security subsystem for policy enforcement, sandboxing, and secret management.
- service
- skillforge
- SkillForge — Skill auto-discovery, evaluation, and integration engine.
- skills
- sop
- subagent
- Runtime-spawned ephemeral sub-agents that inherit their parent
agent’s identity by default: same UUID, same
SecurityPolicy, same memory allowlist. A SubAgent run is auditable as a child of the parent and stays inside the parent’s permissions envelope. - tools
- Tool subsystem for agent-callable capabilities.
- trust
- tunnel
- util
- Utility functions for
ZeroClaw. - verifiable_
intent - Verifiable Intent (VI) — Rust-native implementation of the VI specification.