TaskFlow
DashboardFreewriteWhiteboardsProjectsCRMTasksNotificationsSettingsAgent TowerAPI Docs
OpenClaw Docs
?

User

Member

Caricamento in corso...

Home
Progetti
Task
Notifiche
CRM

    OpenClaw

    Documentation Mirror

    Documentation Overview

    Docs

    Auth credential semantics
    Scheduled tasks
    Hooks
    Automation & tasks
    Standing orders
    Task flow
    Background tasks
    BlueBubbles
    Broadcast groups
    Channel routing
    Discord
    Feishu
    Google Chat
    Group messages
    Groups
    iMessage
    Chat channels
    IRC
    LINE
    Channel location parsing
    Matrix
    Matrix migration
    Matrix push rules for quiet previews
    Mattermost
    Microsoft Teams
    Nextcloud Talk
    Nostr
    Pairing
    QA channel
    QQ bot
    Signal
    Slack
    Synology Chat
    Telegram
    Tlon
    Channel troubleshooting
    Twitch
    WeChat
    WhatsApp
    Yuanbao
    Zalo
    Zalo personal
    CI pipeline
    ACP
    Agent
    Agents
    Approvals
    Backup
    Browser
    Channels
    Clawbot
    `openclaw commitments`
    Completion
    Config
    Configure
    Cron
    Daemon
    Dashboard
    Devices
    Directory
    DNS
    Docs
    Doctor
    Flows (redirect)
    Gateway
    Health
    Hooks
    CLI reference
    Inference CLI
    Logs
    MCP
    Memory
    Message
    Migrate
    Models
    Node
    Nodes
    Onboard
    Pairing
    Plugins
    Proxy
    QR
    Reset
    Sandbox CLI
    Secrets
    Security
    Sessions
    Setup
    Skills
    Status
    System
    `openclaw tasks`
    TUI
    Uninstall
    Update
    Voicecall
    Webhooks
    Wiki
    Active memory
    Agent runtime
    Agent loop
    Agent runtimes
    Agent workspace
    Gateway architecture
    Channel docking
    Inferred commitments
    Compaction
    Context
    Context engine
    Delegate architecture
    Dreaming
    Experimental features
    Features
    Markdown formatting
    Memory overview
    Builtin memory engine
    Honcho memory
    QMD memory engine
    Memory search
    Messages
    Model failover
    Model providers
    Models CLI
    Multi-agent routing
    OAuth
    OpenClaw App SDK
    Presence
    QA overview
    Matrix QA
    Command queue
    Steering queue
    Retry policy
    Session management
    Session pruning
    Session tools
    SOUL.md personality guide
    Streaming and chunking
    System prompt
    Timezones
    TypeBox
    Typing indicators
    Usage tracking
    Date and time
    Node + tsx crash
    Diagnostics flags
    Authentication
    Background exec and process tool
    Bonjour discovery
    Bridge protocol
    CLI backends
    Configuration — agents
    Configuration — channels
    Configuration — tools and custom providers
    Configuration
    Configuration examples
    Configuration reference
    Diagnostics export
    Discovery and transports
    Doctor
    Gateway lock
    Health checks
    Heartbeat
    Gateway runbook
    Local models
    Gateway logging
    Multiple gateways
    Network model
    OpenAI chat completions
    OpenResponses API
    OpenShell
    OpenTelemetry export
    Gateway-owned pairing
    Prometheus metrics
    Gateway protocol
    Remote access
    Remote gateway setup
    Sandbox vs tool policy vs elevated
    Sandboxing
    Secrets management
    Secrets apply plan contract
    Security audit checks
    Security
    Tailscale
    Tools invoke API
    Troubleshooting
    Trusted proxy auth
    Debugging
    Environment variables
    FAQ
    FAQ: first-run setup
    FAQ: models and auth
    GPT-5.5 / Codex agentic parity
    GPT-5.5 / Codex parity maintainer notes
    Help
    Scripts
    Testing
    Testing: live suites
    General troubleshooting
    OpenClaw
    Ansible
    Azure
    Bun (experimental)
    ClawDock
    Release channels
    DigitalOcean
    Docker
    Docker VM runtime
    exe.dev
    Fly.io
    GCP
    Hetzner
    Hostinger
    Install
    Installer internals
    Kubernetes
    macOS VMs
    Migration guide
    Migrating from Claude
    Migrating from Hermes
    Nix
    Node.js
    Northflank
    Oracle Cloud
    Podman
    Railway
    Raspberry Pi
    Render
    Uninstall
    Updating
    Logging
    Network
    Audio and voice notes
    Camera capture
    Image and media support
    Nodes
    Location command
    Media understanding
    Talk mode
    Node troubleshooting
    Voice wake
    Pi integration architecture
    Pi development workflow
    Android app
    Platforms
    iOS app
    Linux app
    Gateway on macOS
    Canvas
    Gateway lifecycle
    macOS dev setup
    Health checks (macOS)
    Menu bar icon
    macOS logging
    Menu bar
    Peekaboo bridge
    macOS permissions
    Remote control
    macOS signing
    Skills (macOS)
    Voice overlay
    Voice wake (macOS)
    WebChat (macOS)
    macOS IPC
    macOS app
    Windows
    Plugin internals
    Plugin architecture internals
    Building plugins
    Plugin bundles
    Codex Computer Use
    Codex harness
    Community plugins
    Plugin compatibility
    Google Meet plugin
    Plugin hooks
    Plugin manifest
    Memory LanceDB
    Memory wiki
    Message presentation
    Agent harness plugins
    Building channel plugins
    Channel turn kernel
    Plugin entry points
    Plugin SDK migration
    Plugin SDK overview
    Building provider plugins
    Plugin runtime helpers
    Plugin setup and config
    Plugin SDK subpaths
    Plugin testing
    Skill workshop plugin
    Voice call plugin
    Webhooks plugin
    Zalo personal plugin
    OpenProse
    Alibaba Model Studio
    Anthropic
    Arcee AI
    Azure Speech
    Amazon Bedrock
    Amazon Bedrock Mantle
    Chutes
    Claude Max API proxy
    Cloudflare AI gateway
    ComfyUI
    Deepgram
    Deepinfra
    DeepSeek
    ElevenLabs
    Fal
    Fireworks
    GitHub Copilot
    GLM (Zhipu)
    Google (Gemini)
    Gradium
    Groq
    Hugging Face (inference)
    Provider directory
    Inferrs
    Inworld
    Kilocode
    LiteLLM
    LM Studio
    MiniMax
    Mistral
    Model provider quickstart
    Moonshot AI
    NVIDIA
    Ollama
    OpenAI
    OpenCode
    OpenCode Go
    OpenRouter
    Perplexity
    Qianfan
    Qwen
    Runway
    SGLang
    StepFun
    Synthetic
    Tencent Cloud (TokenHub)
    Together AI
    Venice AI
    Vercel AI gateway
    vLLM
    Volcengine (Doubao)
    Vydra
    xAI
    Xiaomi MiMo
    Z.AI
    Default AGENTS.md
    Release policy
    API usage and costs
    Credits
    Device model database
    Full release validation
    Memory configuration reference
    OpenClaw App SDK API design
    Prompt caching
    Rich output protocol
    RPC adapters
    SecretRef credential surface
    Session management deep dive
    AGENTS.md template
    BOOT.md template
    BOOTSTRAP.md template
    HEARTBEAT.md template
    IDENTITY template
    SOUL.md template
    TOOLS.md template
    USER template
    Tests
    Token use and costs
    Transcript hygiene
    Onboarding reference
    Contributing to the threat model
    Threat model (MITRE ATLAS)
    Formal verification (security models)
    Network proxy
    Agent bootstrapping
    Docs directory
    Getting started
    Docs hubs
    OpenClaw lore
    Onboarding (macOS app)
    Onboarding overview
    Personal assistant setup
    Setup
    Showcase
    Onboarding (CLI)
    CLI automation
    CLI setup reference
    ACP agents
    ACP agents — setup
    Agent send
    apply_patch tool
    Brave search
    Browser (OpenClaw-managed)
    Browser control API
    Browser troubleshooting
    Browser login
    WSL2 + Windows + remote Chrome CDP troubleshooting
    BTW side questions
    ClawHub
    Code execution
    Creating skills
    Diffs
    DuckDuckGo search
    Elevated mode
    Exa search
    Exec tool
    Exec approvals
    Exec approvals — advanced
    Firecrawl
    Gemini search
    Grok search
    Image generation
    Tools and plugins
    Kimi search
    LLM task
    Lobster
    Tool-loop detection
    Media overview
    MiniMax search
    Multi-agent sandbox and tools
    Music generation
    Ollama web search
    PDF tool
    Perplexity search
    Plugins
    Reactions
    SearXNG search
    Skills
    Skills config
    Slash commands
    Sub-agents
    Tavily
    Thinking levels
    Tokenjuice
    Trajectory bundles
    Text-to-speech
    Video generation
    Web search
    Web fetch
    Linux server
    Control UI
    Dashboard
    Web
    TUI
    WebChat

    OpenAPI Specs

    openapi
    TaskFlow
    docs/openclaw
    Original Docs

    Real-time Synchronized Documentation

    Last sync: 01/05/2026 07:03:07

    Note: This content is mirrored from docs.openclaw.ai and is subject to their terms and conditions.

    OpenClaw Docs

    v2.4.0 Production

    Last synced: Today, 22:00

    Technical reference for the OpenClaw framework. Real-time synchronization with the official documentation engine.

    Use this file to discover all available pages before exploring further.

    ACP agents

    Agent Client Protocol (ACP) sessions let OpenClaw run external coding harnesses (for example Pi, Claude Code, Cursor, Copilot, Droid, OpenClaw ACP, OpenCode, Gemini CLI, and other supported ACPX harnesses) through an ACP backend plugin.

    Each ACP session spawn is tracked as a background task.

    note

    **ACP is the external-harness path, not the default Codex path.** The native Codex app-server plugin owns `/codex ...` controls and the `agentRuntime.id: "codex"` embedded runtime; ACP owns `/acp ...` controls and `sessions_spawn({ runtime: "acp" })` sessions.

    If you want Codex or Claude Code to connect as an external MCP client directly to existing OpenClaw channel conversations, use

    text
    openclaw mcp serve
    instead of ACP.

    Which page do I want?

    You want to…Use thisNotes
    Bind or control Codex in the current conversation
    text
    /codex bind
    ,
    text
    /codex threads
    Native Codex app-server path when the
    text
    codex
    plugin is enabled; includes bound chat replies, image forwarding, model/fast/permissions, stop, and steer controls. ACP is an explicit fallback
    Run Claude Code, Gemini CLI, explicit Codex ACP, or another external harness through OpenClawThis pageChat-bound sessions,
    text
    /acp spawn
    ,
    text
    sessions_spawn({ runtime: "acp" })
    , background tasks, runtime controls
    Expose an OpenClaw Gateway session as an ACP server for an editor or client
    text
    openclaw acp
    Bridge mode. IDE/client talks ACP to OpenClaw over stdio/WebSocket
    Reuse a local AI CLI as a text-only fallback modelCLI BackendsNot ACP. No OpenClaw tools, no ACP controls, no harness runtime

    Does this work out of the box?

    Usually yes. Fresh installs ship the bundled

    text
    acpx
    runtime plugin enabled by default with a plugin-local pinned
    text
    acpx
    binary that OpenClaw probes and self-repairs on startup. Run
    text
    /acp doctor
    for a readiness check.

    OpenClaw only teaches agents about ACP spawning when ACP is truly usable: ACP must be enabled, dispatch must not be disabled, the current session must not be sandbox-blocked, and a runtime backend must be loaded. If those conditions are not met, ACP plugin skills and

    text
    sessions_spawn
    ACP guidance stay hidden so the agent does not suggest an unavailable backend.

    OpenClaw plugin tools and built-in OpenClaw tools are not exposed to ACP harnesses by default. Enable the explicit MCP bridges in ACP agents — setup only when the harness should call those tools directly.

    Supported harness targets

    With the bundled

    text
    acpx
    backend, use these harness ids as
    text
    /acp spawn <id>
    or
    text
    sessions_spawn({ runtime: "acp", agentId: "<id>" })
    targets:

    Harness idTypical backendNotes
    text
    claude
    Claude Code ACP adapterRequires Claude Code auth on the host.
    text
    codex
    Codex ACP adapterExplicit ACP fallback only when native
    text
    /codex
    is unavailable or ACP is requested.
    text
    copilot
    GitHub Copilot ACP adapterRequires Copilot CLI/runtime auth.
    text
    cursor
    Cursor CLI ACP (
    text
    cursor-agent acp
    )
    Override the acpx command if a local install exposes a different ACP entrypoint.
    text
    droid
    Factory Droid CLIRequires Factory/Droid auth or
    text
    FACTORY_API_KEY
    in the harness environment.
    text
    gemini
    Gemini CLI ACP adapterRequires Gemini CLI auth or API key setup.
    text
    iflow
    iFlow CLIAdapter availability and model control depend on the installed CLI.
    text
    kilocode
    Kilo Code CLIAdapter availability and model control depend on the installed CLI.
    text
    kimi
    Kimi/Moonshot CLIRequires Kimi/Moonshot auth on the host.
    text
    kiro
    Kiro CLIAdapter availability and model control depend on the installed CLI.
    text
    opencode
    OpenCode ACP adapterRequires OpenCode CLI/provider auth.
    text
    openclaw
    OpenClaw Gateway bridge through
    text
    openclaw acp
    Lets an ACP-aware harness talk back to an OpenClaw Gateway session.
    text
    pi
    Pi/embedded OpenClaw runtimeUsed for OpenClaw-native harness experiments.
    text
    qwen
    Qwen Code / Qwen CLIRequires Qwen-compatible auth on the host.

    Custom acpx agent aliases can be configured in acpx itself, but OpenClaw policy still checks

    text
    acp.allowedAgents
    and any
    text
    agents.list[].runtime.acp.agent
    mapping before dispatch.

    Operator runbook

    Quick

    text
    /acp
    flow from chat:

    Spawn

    `/acp spawn claude --bind here`, `/acp spawn gemini --mode persistent --thread auto`, or explicit `/acp spawn codex --bind here`.

    Work

    Continue in the bound conversation or thread (or target the session key explicitly).

    Check state

    `/acp status`

    Tune

    `/acp model `, `/acp permissions `, `/acp timeout `.

    Steer

    Without replacing context: `/acp steer tighten logging and continue`.

    Stop

    `/acp cancel` (current turn) or `/acp close` (session + bindings).

    ACP versus sub-agents

    Use ACP when you want an external harness runtime. Use native Codex app-server for Codex conversation binding/control when the

    text
    codex
    plugin is enabled. Use sub-agents when you want OpenClaw-native delegated runs.

    AreaACP sessionSub-agent run
    RuntimeACP backend plugin (for example acpx)OpenClaw native sub-agent runtime
    Session key
    text
    agent:<agentId>:acp:<uuid>
    text
    agent:<agentId>:subagent:<uuid>
    Main commands
    text
    /acp ...
    text
    /subagents ...
    Spawn tool
    text
    sessions_spawn
    with
    text
    runtime:"acp"
    text
    sessions_spawn
    (default runtime)

    See also Sub-agents.

    How ACP runs Claude Code

    For Claude Code through ACP, the stack is:

    1. OpenClaw ACP session control plane.
    2. Bundled
      text
      acpx
      runtime plugin.
    3. Claude ACP adapter.
    4. Claude-side runtime/session machinery.

    ACP Claude is a harness session with ACP controls, session resume, background-task tracking, and optional conversation/thread binding.

    CLI backends are separate text-only local fallback runtimes — see CLI Backends.

    For operators, the practical rule is:

    • Want
      text
      /acp spawn
      , bindable sessions, runtime controls, or persistent harness work?
      Use ACP.
    • Want simple local text fallback through the raw CLI? Use CLI backends.

    Bound sessions

    Mental model

    • Chat surface — where people keep talking (Discord channel, Telegram topic, iMessage chat).
    • ACP session — the durable Codex/Claude/Gemini runtime state OpenClaw routes to.
    • Child thread/topic — an optional extra messaging surface created only by
      text
      --thread ...
      .
    • Runtime workspace — the filesystem location (
      text
      cwd
      , repo checkout, backend workspace) where the harness runs. Independent of the chat surface.

    Current-conversation binds

    text
    /acp spawn <harness> --bind here
    pins the current conversation to the spawned ACP session — no child thread, same chat surface. OpenClaw keeps owning transport, auth, safety, and delivery. Follow-up messages in that conversation route to the same session;
    text
    /new
    and
    text
    /reset
    reset the session in place;
    text
    /acp close
    removes the binding.

    Examples:

    text
    /codex bind # native Codex bind, route future messages here /codex model gpt-5.4 # tune the bound native Codex thread /codex stop # control the active native Codex turn /acp spawn codex --bind here # explicit ACP fallback for Codex /acp spawn codex --thread auto # may create a child thread/topic and bind there /acp spawn codex --bind here --cwd /workspace/repo # same chat binding, Codex runs in /workspace/repo

    Persistent channel bindings

    For non-ephemeral workflows, configure persistent ACP bindings in top-level

    text
    bindings[]
    entries.

    Binding model

    Marks a persistent ACP conversation binding. Identifies the target conversation. Per-channel shapes:
    • Discord channel/thread:
      text
      match.channel="discord"
      +
      text
      match.peer.id="<channelOrThreadId>"
    • Telegram forum topic:
      text
      match.channel="telegram"
      +
      text
      match.peer.id="<chatId>:topic:<topicId>"
    • BlueBubbles DM/group:
      text
      match.channel="bluebubbles"
      +
      text
      match.peer.id="<handle|chat_id:*|chat_guid:*|chat_identifier:*>"
      . Prefer
      text
      chat_id:*
      or
      text
      chat_identifier:*
      for stable group bindings.
    • iMessage DM/group:
      text
      match.channel="imessage"
      +
      text
      match.peer.id="<handle|chat_id:*|chat_guid:*|chat_identifier:*>"
      . Prefer
      text
      chat_id:*
      for stable group bindings.
    The owning OpenClaw agent id. Optional ACP override. Optional operator-facing label. Optional runtime working directory. Optional backend override.

    Runtime defaults per agent

    Use

    text
    agents.list[].runtime
    to define ACP defaults once per agent:

    • text
      agents.list[].runtime.type="acp"
    • text
      agents.list[].runtime.acp.agent
      (harness id, e.g.
      text
      codex
      or
      text
      claude
      )
    • text
      agents.list[].runtime.acp.backend
    • text
      agents.list[].runtime.acp.mode
    • text
      agents.list[].runtime.acp.cwd

    Override precedence for ACP bound sessions:

    1. text
      bindings[].acp.*
    2. text
      agents.list[].runtime.acp.*
    3. Global ACP defaults (e.g.
      text
      acp.backend
      )

    Example

    json5
    { agents: { list: [ { id: "codex", runtime: { type: "acp", acp: { agent: "codex", backend: "acpx", mode: "persistent", cwd: "/workspace/openclaw", }, }, }, { id: "claude", runtime: { type: "acp", acp: { agent: "claude", backend: "acpx", mode: "persistent" }, }, }, ], }, bindings: [ { type: "acp", agentId: "codex", match: { channel: "discord", accountId: "default", peer: { kind: "channel", id: "222222222222222222" }, }, acp: { label: "codex-main" }, }, { type: "acp", agentId: "claude", match: { channel: "telegram", accountId: "default", peer: { kind: "group", id: "-1001234567890:topic:42" }, }, acp: { cwd: "/workspace/repo-b" }, }, { type: "route", agentId: "main", match: { channel: "discord", accountId: "default" }, }, { type: "route", agentId: "main", match: { channel: "telegram", accountId: "default" }, }, ], channels: { discord: { guilds: { "111111111111111111": { channels: { "222222222222222222": { requireMention: false }, }, }, }, }, telegram: { groups: { "-1001234567890": { topics: { "42": { requireMention: false } }, }, }, }, }, }

    Behavior

    • OpenClaw ensures the configured ACP session exists before use.
    • Messages in that channel or topic route to the configured ACP session.
    • In bound conversations,
      text
      /new
      and
      text
      /reset
      reset the same ACP session key in place.
    • Temporary runtime bindings (for example created by thread-focus flows) still apply where present.
    • For cross-agent ACP spawns without an explicit
      text
      cwd
      , OpenClaw inherits the target agent workspace from agent config.
    • Missing inherited workspace paths fall back to the backend default cwd; non-missing access failures surface as spawn errors.

    Start ACP sessions

    Two ways to start an ACP session:

    Use `runtime: "acp"` to start an ACP session from an agent turn or tool call.
    text
    ```json} { "task": "Open the repo and summarize failing tests", "runtime": "acp", "agentId": "codex", "thread": true, "mode": "session" } ``` <Note> `runtime` defaults to `subagent`, so set `runtime: "acp"` explicitly for ACP sessions. If `agentId` is omitted, OpenClaw uses `acp.defaultAgent` when configured. `mode: "session"` requires `thread: true` to keep a persistent bound conversation. </Note>
    Use `/acp spawn` for explicit operator control from chat.
    text
    ```text} /acp spawn codex --mode persistent --thread auto /acp spawn codex --mode oneshot --thread off /acp spawn codex --bind here /acp spawn codex --thread here ``` Key flags: * `--mode persistent|oneshot` * `--bind here|off` * `--thread auto|here|off` * `--cwd <absolute-path>` * `--label <name>` See [Slash commands](/tools/slash-commands).

    text
    sessions_spawn
    parameters

    Initial prompt sent to the ACP session. Must be `"acp"` for ACP sessions. ACP target harness id. Falls back to `acp.defaultAgent` if set. Request thread binding flow where supported. `"run"` is one-shot; `"session"` is persistent. If `thread: true` and `mode` is omitted, OpenClaw may default to persistent behaviour per runtime path. `mode: "session"` requires `thread: true`. Requested runtime working directory (validated by backend/runtime policy). If omitted, ACP spawn inherits the target agent workspace when configured; missing inherited paths fall back to backend defaults, while real access errors are returned. Operator-facing label used in session/banner text. Resume an existing ACP session instead of creating a new one. The agent replays its conversation history via `session/load`. Requires `runtime: "acp"`. `"parent"` streams initial ACP run progress summaries back to the requester session as system events. Accepted responses include `streamLogPath` pointing to a session-scoped JSONL log (`.acp-stream.jsonl`) you can tail for full relay history. Aborts the ACP child turn after N seconds. `0` keeps the turn on the gateway's no-timeout path. The same value is applied to the Gateway run and ACP runtime so stalled/quota-exhausted harnesses do not occupy the parent agent lane indefinitely. Explicit model override for the ACP child session. Codex ACP spawns normalize OpenClaw Codex refs such as `openai-codex/gpt-5.4` to Codex ACP startup config before `session/new`; slash forms such as `openai-codex/gpt-5.4/high` also set Codex ACP reasoning effort. Other harnesses must advertise ACP `models` and support `session/set_model`; otherwise OpenClaw/acpx fails clearly instead of silently falling back to the target agent default. Explicit thinking/reasoning effort. For Codex ACP, `minimal` maps to low effort, `low`/`medium`/`high`/`xhigh` map directly, and `off` omits the reasoning-effort startup override.

    Spawn bind and thread modes

    | Mode | Behavior | | ------ | ---------------------------------------------------------------------- | | `here` | Bind the current active conversation in place; fail if none is active. | | `off` | Do not create a current-conversation binding. |
    text
    Notes: * `--bind here` is the simplest operator path for "make this channel or chat Codex-backed." * `--bind here` does not create a child thread. * `--bind here` is only available on channels that expose current-conversation binding support. * `--bind` and `--thread` cannot be combined in the same `/acp spawn` call.
    | Mode | Behavior | | ------ | --------------------------------------------------------------------------------------------------- | | `auto` | In an active thread: bind that thread. Outside a thread: create/bind a child thread when supported. | | `here` | Require current active thread; fail if not in one. | | `off` | No binding. Session starts unbound. |
    text
    Notes: * On non-thread binding surfaces, default behavior is effectively `off`. * Thread-bound spawn requires channel policy support: * Discord: `channels.discord.threadBindings.spawnAcpSessions=true` * Telegram: `channels.telegram.threadBindings.spawnAcpSessions=true` * Use `--bind here` when you want to pin the current conversation without creating a child thread.

    Delivery model

    ACP sessions can be either interactive workspaces or parent-owned background work. The delivery path depends on that shape.

    Sandbox compatibility

    ACP sessions currently run on the host runtime, not inside the OpenClaw sandbox.

    warning

    **Security boundary:**
    • The external harness can read/write according to its own CLI permissions and the selected
      text
      cwd
      .
    • OpenClaw's sandbox policy does not wrap ACP harness execution.
    • OpenClaw still enforces ACP feature gates, allowed agents, session ownership, channel bindings, and Gateway delivery policy.
    • Use
      text
      runtime: "subagent"
      for sandbox-enforced OpenClaw-native work.

    Current limitations:

    • If the requester session is sandboxed, ACP spawns are blocked for both
      text
      sessions_spawn({ runtime: "acp" })
      and
      text
      /acp spawn
      .
    • text
      sessions_spawn
      with
      text
      runtime: "acp"
      does not support
      text
      sandbox: "require"
      .

    Session target resolution

    Most

    text
    /acp
    actions accept an optional session target (
    text
    session-key
    ,
    text
    session-id
    , or
    text
    session-label
    ).

    Resolution order:

    1. Explicit target argument (or
      text
      --session
      for
      text
      /acp steer
      )
      • tries key
      • then UUID-shaped session id
      • then label
    2. Current thread binding (if this conversation/thread is bound to an ACP session).
    3. Current requester session fallback.

    Current-conversation bindings and thread bindings both participate in step 2.

    If no target resolves, OpenClaw returns a clear error (

    text
    Unable to resolve session target: ...
    ).

    ACP controls

    CommandWhat it doesExample
    text
    /acp spawn
    Create ACP session; optional current bind or thread bind.
    text
    /acp spawn codex --bind here --cwd /repo
    text
    /acp cancel
    Cancel in-flight turn for target session.
    text
    /acp cancel agent:codex:acp:<uuid>
    text
    /acp steer
    Send steer instruction to running session.
    text
    /acp steer --session support inbox prioritize failing tests
    text
    /acp close
    Close session and unbind thread targets.
    text
    /acp close
    text
    /acp status
    Show backend, mode, state, runtime options, capabilities.
    text
    /acp status
    text
    /acp set-mode
    Set runtime mode for target session.
    text
    /acp set-mode plan
    text
    /acp set
    Generic runtime config option write.
    text
    /acp set model openai/gpt-5.4
    text
    /acp cwd
    Set runtime working directory override.
    text
    /acp cwd /Users/user/Projects/repo
    text
    /acp permissions
    Set approval policy profile.
    text
    /acp permissions strict
    text
    /acp timeout
    Set runtime timeout (seconds).
    text
    /acp timeout 120
    text
    /acp model
    Set runtime model override.
    text
    /acp model anthropic/claude-opus-4-6
    text
    /acp reset-options
    Remove session runtime option overrides.
    text
    /acp reset-options
    text
    /acp sessions
    List recent ACP sessions from store.
    text
    /acp sessions
    text
    /acp doctor
    Backend health, capabilities, actionable fixes.
    text
    /acp doctor
    text
    /acp install
    Print deterministic install and enable steps.
    text
    /acp install

    text
    /acp status
    shows the effective runtime options plus runtime-level and backend-level session identifiers. Unsupported-control errors surface clearly when a backend lacks a capability.
    text
    /acp sessions
    reads the store for the current bound or requester session; target tokens (
    text
    session-key
    ,
    text
    session-id
    , or
    text
    session-label
    ) resolve through gateway session discovery, including custom per-agent
    text
    session.store
    roots.

    Runtime options mapping

    text
    /acp
    has convenience commands and a generic setter. Equivalent operations:

    CommandMaps toNotes
    text
    /acp model <id>
    runtime config key
    text
    model
    For Codex ACP, OpenClaw normalizes
    text
    openai-codex/<model>
    to the adapter model id and maps slash reasoning suffixes such as
    text
    openai-codex/gpt-5.4/high
    to
    text
    reasoning_effort
    .
    text
    /acp set thinking <level>
    runtime config key
    text
    thinking
    For Codex ACP, OpenClaw sends the corresponding
    text
    reasoning_effort
    where the adapter supports one.
    text
    /acp permissions <profile>
    runtime config key
    text
    approval_policy
    —
    text
    /acp timeout <seconds>
    runtime config key
    text
    timeout
    —
    text
    /acp cwd <path>
    runtime cwd overrideDirect update.
    text
    /acp set <key> <value>
    generic
    text
    key=cwd
    uses the cwd override path.
    text
    /acp reset-options
    clears all runtime overrides—

    acpx harness, plugin setup, and permissions

    For acpx harness configuration (Claude Code / Codex / Gemini CLI aliases), the plugin-tools and OpenClaw-tools MCP bridges, and ACP permission modes, see ACP agents — setup.

    Troubleshooting

    SymptomLikely causeFix
    text
    ACP runtime backend is not configured
    Backend plugin missing, disabled, or blocked by
    text
    plugins.allow
    .
    Install and enable backend plugin, include
    text
    acpx
    in
    text
    plugins.allow
    when that allowlist is set, then run
    text
    /acp doctor
    .
    text
    ACP is disabled by policy (acp.enabled=false)
    ACP globally disabled.Set
    text
    acp.enabled=true
    .
    text
    ACP dispatch is disabled by policy (acp.dispatch.enabled=false)
    Automatic dispatch from normal thread messages disabled.Set
    text
    acp.dispatch.enabled=true
    to resume automatic thread routing; explicit
    text
    sessions_spawn({ runtime: "acp" })
    calls still work.
    text
    ACP agent "<id>" is not allowed by policy
    Agent not in allowlist.Use allowed
    text
    agentId
    or update
    text
    acp.allowedAgents
    .
    text
    /acp doctor
    reports backend not ready right after startup
    Plugin dependency probe or self-repair is still running.Wait briefly and rerun
    text
    /acp doctor
    ; if it stays unhealthy, inspect the backend install error and plugin allow/deny policy.
    Harness command not foundAdapter CLI is not installed, staged plugin deps are missing, or first-run
    text
    npx
    fetch failed for a non-Codex adapter.
    Run
    text
    /acp doctor
    , repair plugin dependencies, install/prewarm the adapter on the Gateway host, or configure the acpx agent command explicitly.
    Model-not-found from the harnessModel id is valid for another provider/harness but not this ACP target.Use a model listed by that harness, configure the model in the harness, or omit the override.
    Vendor auth error from the harnessOpenClaw is healthy, but the target CLI/provider is not logged in.Log in or provide the required provider key on the Gateway host environment.
    text
    Unable to resolve session target: ...
    Bad key/id/label token.Run
    text
    /acp sessions
    , copy exact key/label, retry.
    text
    --bind here requires running /acp spawn inside an active ... conversation
    text
    --bind here
    used without an active bindable conversation.
    Move to the target chat/channel and retry, or use unbound spawn.
    text
    Conversation bindings are unavailable for <channel>.
    Adapter lacks current-conversation ACP binding capability.Use
    text
    /acp spawn ... --thread ...
    where supported, configure top-level
    text
    bindings[]
    , or move to a supported channel.
    text
    --thread here requires running /acp spawn inside an active ... thread
    text
    --thread here
    used outside a thread context.
    Move to target thread or use
    text
    --thread auto
    /
    text
    off
    .
    text
    Only <user-id> can rebind this channel/conversation/thread.
    Another user owns the active binding target.Rebind as owner or use a different conversation or thread.
    text
    Thread bindings are unavailable for <channel>.
    Adapter lacks thread binding capability.Use
    text
    --thread off
    or move to supported adapter/channel.
    text
    Sandboxed sessions cannot spawn ACP sessions ...
    ACP runtime is host-side; requester session is sandboxed.Use
    text
    runtime="subagent"
    from sandboxed sessions, or run ACP spawn from a non-sandboxed session.
    text
    sessions_spawn sandbox="require" is unsupported for runtime="acp" ...
    text
    sandbox="require"
    requested for ACP runtime.
    Use
    text
    runtime="subagent"
    for required sandboxing, or use ACP with
    text
    sandbox="inherit"
    from a non-sandboxed session.
    text
    Cannot apply --model ... did not advertise model support
    The target harness does not expose generic ACP model switching.Use a harness that advertises ACP
    text
    models
    /
    text
    session/set_model
    , use Codex ACP model refs, or configure the model directly in the harness if it has its own startup flag.
    Missing ACP metadata for bound sessionStale/deleted ACP session metadata.Recreate with
    text
    /acp spawn
    , then rebind/focus thread.
    text
    AcpRuntimeError: Permission prompt unavailable in non-interactive mode
    text
    permissionMode
    blocks writes/exec in non-interactive ACP session.
    Set
    text
    plugins.entries.acpx.config.permissionMode
    to
    text
    approve-all
    and restart gateway. See Permission configuration.
    ACP session fails early with little outputPermission prompts are blocked by
    text
    permissionMode
    /
    text
    nonInteractivePermissions
    .
    Check gateway logs for
    text
    AcpRuntimeError
    . For full permissions, set
    text
    permissionMode=approve-all
    ; for graceful degradation, set
    text
    nonInteractivePermissions=deny
    .
    ACP session stalls indefinitely after completing workHarness process finished but ACP session did not report completion.Monitor with
    text
    ps aux | grep acpx
    ; kill stale processes manually.
    Harness sees
    text
    <<<BEGIN_OPENCLAW_INTERNAL_CONTEXT>>>
    Internal event envelope leaked across the ACP boundary.Update OpenClaw and rerun the completion flow; external harnesses should receive plain completion prompts only.

    Related

    • ACP agents — setup
    • Agent send
    • CLI Backends
    • Codex harness
    • Multi-agent sandbox tools
    • text
      openclaw acp
      (bridge mode)
    • Sub-agents

    © 2024 TaskFlow Mirror

    Powered by TaskFlow Sync Engine