OpenClaw Reference: Pi Integration & Configuration Reference

🌏 中文版

OpenClaw’s AI core is Pi (a coding agent runtime). OpenClaw handles channels, routing, and security, while Pi handles reasoning, tool execution, and memory. This article summarizes the relationship between the two and provides a configuration reference.

Relationship Between OpenClaw and Pi

Layer	Owner	Function
Gateway	OpenClaw	Channel management, routing, authentication, scheduling
Agent Runtime	Pi	Reasoning, tool execution, context management
Session	Co-managed	OpenClaw owns the session; Pi executes the agent loop
Memory	Pi	Markdown files, vector search
Compaction	Pi	Context window compression
Sandbox	Co-managed	OpenClaw manages containers; Pi executes inside them

Pi’s configuration is passed through OpenClaw’s agents.defaults and agents.list[].

Complete Feature List

Core

Feature	Description
Multi-agent routing	Multiple agent routing (binding priority order)
Delegate architecture	3-tier autonomous authorization
Agent loop	intake → context → inference → tool → stream
Block streaming	Chunking (minChars/maxChars)
5 queue modes	steer/followup/collect/steer-backlog/interrupt

Models

Feature	Description
35+ providers	Anthropic/OpenAI/Google + local (Ollama/vLLM)
Auth profile rotation	Cooldown escalation
Model failover	Auth → Model → Thinking degradation
Prompt caching	cacheRetention (none/short/long)
Token tracking	10 cost source types

Session & Memory

Feature	Description
4 DM scopes	main/per-peer/per-channel-peer/per-account-channel-peer
Identity links	Cross-platform same-person linking
Memory flush	Auto-write to disk before compaction
Vector search	BM25 + embedding hybrid search
Context Engine	Pluggable (ingest/assemble/compact/after-turn)

Security

Feature	Description
3 sandbox backends	Docker/SSH/OpenShell
3 control layers	Sandbox/Tool Policy/Elevated
MITRE ATLAS threat model	13 threat scenarios
TLA+ formal verification	6+ security assertions
SecretRef	env/file/exec — three source types

Channels

Feature	Description
24+ channels	Running simultaneously
Pairing	8-character code, 1-hour expiry
Broadcast Groups	Multiple agents processing concurrently
Channel routing	Deterministic reply to source

Tools

Feature	Description
Browser	Managed profile + remote CDP
Exec	3 security levels + approval
Skills	6-tier priority + ClawHub
Sub-agent	Maximum 5 nesting levels
Lobster	Deterministic workflow runtime
TTS	3 providers
PDF	Native + extraction

Automation

Feature	Description
Cron	Precise scheduling + isolated sessions
Heartbeat	Periodic health checks
Webhook	External event triggers
Standing Orders	Persistent authorization programs

UI & Nodes

Feature	Description
Control UI	Browser dashboard
TUI	Terminal interactive interface
Web Chat	WebSocket real-time chat
iOS/Android Node	Camera/Canvas/Location/SMS
Node Host	Remote exec

Configuration Structure Quick Reference

{
  agents: {           // Agent settings (defaults + list)
    defaults: {
      workspace: "",
      sandbox: {},
      heartbeat: {},
      compaction: {},
      models: {}
    },
    list: []
  },
  channels: {},       // Channel settings (per-channel)
  commands: {},       // Slash commands settings
  gateway: {          // Gateway settings
    bind: "",
    port: 0,
    auth: {},
    tailscale: {},
    controlUi: {}
  },
  hooks: {},          // Webhook settings
  mcp: {},            // MCP Server settings
  messages: {},       // Message settings (TTS, formatting)
  models: {},         // Model provider settings
  plugins: {},        // Plugin settings
  secrets: {},        // SecretRef provider settings
  session: {},        // Session settings (dmScope, maintenance)
  skills: {},         // Skills settings
  tools: {}           // Tool settings (exec, browser, web search)
}

CLI Quick Reference

Gateway

openclaw gateway              # Start Gateway
openclaw health               # Health check
openclaw doctor               # Diagnostics
openclaw doctor --fix         # Auto-fix

Configuration

openclaw config get <path>    # Read
openclaw config set <path> <value>  # Set
openclaw config validate      # Validate
openclaw onboard              # Interactive setup

Models

openclaw models status        # Auth status
openclaw models auth login    # Login
openclaw models auth setup-token  # Setup token

Channels

openclaw channels status      # Channel status
openclaw channels login       # Channel login
openclaw channels add         # Add channel
openclaw pairing list         # Pairing list
openclaw pairing approve      # Approve pairing

Automation

openclaw cron add             # Add cron
openclaw cron list            # List cron jobs
openclaw cron remove          # Remove cron

Nodes

openclaw devices list         # List devices
openclaw devices approve      # Approve device
openclaw nodes status         # Node status
openclaw node run             # Start node host

Sandbox

openclaw sandbox explain      # View settings
openclaw sandbox recreate     # Recreate sandbox
openclaw sandbox list         # List sandboxes

Plugin

openclaw plugins install      # Install
openclaw plugins list         # List
openclaw plugins status       # Status

Overall

This article serves as the reference index for the entire series. OpenClaw’s feature coverage is extensive — from 24+ channels to 35+ model providers, from sandboxing to formal verification, from a Skills marketplace to a deterministic workflow engine. Most users will only need about 20% of these features, but knowing what’s available is important.

References

This article is compiled from the following OpenClaw source documents:

• docs/reference/pi.md — Pi integration
• docs/concepts/features.md — Feature list
• docs/concepts/architecture.md — Architecture
• docs/gateway/configuration-reference.md — Configuration reference
• docs/cli/index.md — CLI overview