English简体中文繁體中文日本語EspañolPortuguês (BR)한국어DeutschFrançaisالعربيةItalianoTiếng ViệtNederlandsTürkçeУкраїнськаBahasa IndonesiaPolskiفارسیไทย
View as MarkdownView this page as plain textOpen in ChatGPTAsk questions about this pageOpen in ClaudeAsk questions about this pageOpen in PerplexityAsk questions about this page

Agent coordination

بيئة معزولة وأدوات متعددة الوكلاء

Status: active

Each agent in a multi-agent setup can override the global sandbox and tool policy. This page covers per-agent configuration, precedence rules, and examples.

Sandboxing

Backends and modes — full sandbox reference.
Sandbox vs tool policy vs elevated

Debug "why is this blocked?"
Elevated mode

Elevated exec for trusted senders.

Configuration examples

Example 1: Personal + restricted family agent
json
{  "agents": {    "list": [      {        "id": "main",        "default": true,        "name": "Personal Assistant",        "workspace": "~/.openclaw/workspace",        "sandbox": { "mode": "off" }      },      {        "id": "family",        "name": "Family Bot",        "workspace": "~/.openclaw/workspace-family",        "sandbox": {          "mode": "all",          "scope": "agent"        },        "tools": {          "allow": ["read", "message"],          "deny": ["exec", "write", "edit", "apply_patch", "process", "browser"],          "message": {            "crossContext": {              "allowWithinProvider": false,              "allowAcrossProviders": false            }          }        }      }    ]  },  "bindings": [    {      "agentId": "family",      "match": {        "provider": "whatsapp",        "accountId": "*",        "peer": {          "kind": "group",          "id": "[email protected]"        }      }    }  ]}

Result:

  • main agent: runs on host, full tool access.
  • family agent: runs in Docker (one container per agent), only read and current-conversation message sends.
Example 2: Work agent with shared sandbox
json
{  "agents": {    "list": [      {        "id": "personal",        "workspace": "~/.openclaw/workspace-personal",        "sandbox": { "mode": "off" }      },      {        "id": "work",        "workspace": "~/.openclaw/workspace-work",        "sandbox": {          "mode": "all",          "scope": "shared",          "workspaceRoot": "/tmp/work-sandboxes"        },        "tools": {          "allow": ["read", "write", "apply_patch", "exec"],          "deny": ["browser", "gateway", "discord"]        }      }    ]  }}
Example 2b: Global coding profile + messaging-only agent
json
{  "tools": { "profile": "coding" },  "agents": {    "list": [      {        "id": "support",        "tools": { "profile": "messaging", "allow": ["slack"] }      }    ]  }}

Result:

  • default agents get coding tools.
  • support agent is messaging-only (+ Slack tool).
Example 3: Different sandbox modes per agent
json
{  "agents": {    "defaults": {      "sandbox": {        "mode": "non-main",        "scope": "session"      }    },    "list": [      {        "id": "main",        "workspace": "~/.openclaw/workspace",        "sandbox": {          "mode": "off"        }      },      {        "id": "public",        "workspace": "~/.openclaw/workspace-public",        "sandbox": {          "mode": "all",          "scope": "agent"        },        "tools": {          "allow": ["read"],          "deny": ["exec", "write", "edit", "apply_patch"]        }      }    ]  }}

Configuration precedence

When both global (agents.defaults.*) and agent-specific (agents.list[].*) configs exist:

Sandbox config

Agent-specific settings override global:

Code
agents.list[].sandbox.mode > agents.defaults.sandbox.modeagents.list[].sandbox.scope > agents.defaults.sandbox.scopeagents.list[].sandbox.workspaceRoot > agents.defaults.sandbox.workspaceRootagents.list[].sandbox.workspaceAccess > agents.defaults.sandbox.workspaceAccessagents.list[].sandbox.docker.* > agents.defaults.sandbox.docker.*agents.list[].sandbox.browser.* > agents.defaults.sandbox.browser.*agents.list[].sandbox.prune.* > agents.defaults.sandbox.prune.*

Tool restrictions

The filtering order is:

  • Tool profile

    tools.profile or agents.list[].tools.profile.

  • Provider tool profile

    tools.byProvider[provider].profile or agents.list[].tools.byProvider[provider].profile.

  • Global tool policy

    tools.allow / tools.deny.

  • Provider tool policy

    tools.byProvider[provider].allow/deny.

  • Agent-specific tool policy

    agents.list[].tools.allow/deny.

  • Agent provider policy

    agents.list[].tools.byProvider[provider].allow/deny.

  • Sandbox tool policy

    tools.sandbox.tools or agents.list[].tools.sandbox.tools.

  • Subagent tool policy

    tools.subagents.tools, if applicable.

    • Precedence rules
    • Each level can further restrict tools, but cannot grant back denied tools from earlier levels.
    • If agents.list[].tools.sandbox.tools is set, it replaces tools.sandbox.tools for that agent.
    • If agents.list[].tools.profile is set, it overrides tools.profile for that agent.
    • Provider tool keys accept either provider (e.g. google-antigravity) or provider/model (e.g. openai/gpt-5.4).
    Empty allowlist behavior

    If any explicit allowlist in that chain leaves the run with no callable tools, OpenClaw stops before submitting the prompt to the model. This is intentional: an agent configured with a missing tool such as agents.list[].tools.allow: ["query_db"] should fail loudly until the plugin that registers query_db is enabled, not continue as a text-only agent.

    Tool policies support group:* shorthands that expand to multiple tools. See Tool groups for the full list.

    Per-agent elevated overrides (agents.list[].tools.elevated) can further restrict elevated exec for specific agents. See Elevated mode for details.

    الترحيل من وكيل واحد

    قبل (وكيل واحد)

    json
    {  "agents": {    "defaults": {      "workspace": "~/.openclaw/workspace",      "sandbox": {        "mode": "non-main"      }    }  },  "tools": {    "sandbox": {      "tools": {        "allow": ["read", "write", "apply_patch", "exec"],        "deny": []      }    }  }}

    بعد (وكلاء متعددون)

    json
    {  "agents": {    "list": [      {        "id": "main",        "default": true,        "workspace": "~/.openclaw/workspace",        "sandbox": { "mode": "off" }      }    ]  }}

    أمثلة تقييد الأدوات

    وكيل للقراءة فقط

    json
    {  "tools": {    "allow": ["read"],    "deny": ["exec", "write", "edit", "apply_patch", "process"]  }}

    تنفيذ Shell مع تعطيل أدوات نظام الملفات

    json
    {  "tools": {    "allow": ["read", "exec", "process"],    "deny": ["write", "edit", "apply_patch", "browser", "gateway"]  }}

    الاتصال فقط

    json
    {  "tools": {    "sessions": { "visibility": "tree" },    "allow": ["sessions_list", "sessions_send", "sessions_history", "session_status"],    "deny": ["exec", "write", "edit", "apply_patch", "read", "browser"]  }}

    لا يزال sessions_history في هذا الملف الشخصي يعيد عرض استدعاء محدودًا ومنقحًا بدلًا من تفريغ نصي خام. يزيل استدعاء المساعد وسوم التفكير، وبنية <relevant-memories>، وحمولات XML لاستدعاءات الأدوات بنص عادي (بما في ذلك <tool_call>...</tool_call> و<function_call>...</function_call> و<tool_calls>...</tool_calls> و<function_calls>...</function_calls> وكتل استدعاء الأدوات المقتطعة)، وبنية استدعاءات الأدوات المخفّضة، ورموز تحكم النموذج المسرّبة بنمط ASCII/العرض الكامل، وXML استدعاءات أدوات MiniMax غير الصالح قبل التنقيح/الاقتطاع.

    خطأ شائع: "non-main"

    الاختبار

    بعد تكوين صندوق العزل والأدوات للوكلاء المتعددين:

  • تحقق من حلّ الوكيل

    bash
    openclaw agents list --bindings

  • تحقق من حاويات صندوق العزل

    bash
    docker ps --filter "name=openclaw-sbx-"

  • اختبر قيود الأدوات

    • أرسل رسالة تتطلب أدوات مقيدة.
    • تحقق من أن الوكيل لا يستطيع استخدام الأدوات الممنوعة.

  • راقب السجلات

    bash
    tail -f "${OPENCLAW_STATE_DIR:-$HOME/.openclaw}/logs/gateway.log" | grep -E "routing|sandbox|tools"

    • استكشاف الأخطاء وإصلاحها

    الوكيل غير موضوع في صندوق العزل رغم `mode: 'all'`
    • تحقق مما إذا كان هناك agents.defaults.sandbox.mode عام يتجاوزه.
    • تكون للإعدادات الخاصة بالوكيل الأولوية، لذلك اضبط agents.list[].sandbox.mode: "all".
    الأدوات لا تزال متاحة رغم قائمة المنع
    • تحقق من ترتيب تصفية الأدوات: عام → وكيل → صندوق عزل → وكيل فرعي.
    • يمكن لكل مستوى أن يزيد التقييد فقط، ولا يمكنه إعادة المنح.
    • تحقق من ذلك عبر السجلات: [tools] filtering tools for agent:${agentId}.
    الحاوية غير معزولة لكل وكيل
    • اضبط scope: "agent" في تكوين صندوق العزل الخاص بالوكيل.
    • الافتراضي هو "session"، وهو ما ينشئ حاوية واحدة لكل جلسة.

    ذات صلة

    Was this useful?