返回
Openclaw-Series

10. OpenClaw 多 Agent 路由与会话隔离

本文深入讲解 OpenClaw 的多 Agent 架构，包括如何配置多个 Agent 实例、消息路由规则、会话隔离策略、以及团队协作场景的最佳实践，帮助你构建企业级的 AI 助手系统。

适用于 OpenClaw v2026.2 | 本文适合团队使用或需要隔离上下文的用户。

TL;DR: 多 Agent 配置："agents": {"instances": {"personal": {...}, "work": {...}}}。路由规则："routing": {"rules": [{"match": {"channel": "telegram"}, "agent": "personal"}]}。会话隔离："sessions": {"isolation": "user"}。沙箱模式："sandbox": {"mode": "non-main"}。适合多用户、多项目、多用途场景。

为什么需要多 Agent？

单 Agent 的局限

场景 单 Agent 问题 多 Agent 解决方案
多用户 上下文混乱，隐私问题 每用户独立 Agent
多项目 项目信息混在一起 每项目独立 Agent
多用途 一个 Agent 多种角色 专门化的 Agent
多团队 权限难以控制 按团队隔离 Agent
成本控制 无法区分用量 每个模型独立计费

多 Agent 使用场景

flowchart TB
    subgraph Users["用户"]
        U1[用户A]
        U2[用户B]
        U3[团队X]
        U4[团队Y]
    end
    
    subgraph Router["消息路由"]
        R[路由规则]
    end
    
    subgraph Agents["Agent 实例"]
        A1[个人Agent A]
        A2[个人Agent B]
        A3[团队Agent X]
        A4[团队Agent Y]
    end
    
    subgraph Models["模型"]
        M1[Claude Opus]
        M2[Claude Sonnet]
        M3[GPT-4o]
        M4[Haiku]
    end
    
    U1 --> R
    U2 --> R
    U3 --> R
    U4 --> R
    R --> A1
    R --> A2
    R --> A3
    R --> A4
    A1 --> M1
    A2 --> M2
    A3 --> M3
    A4 --> M4

多 Agent 配置

基本结构

{
  "agents": {
    "defaults": {
      "model": "anthropic/claude-sonnet-4",
      "workspace": "~/.openclaw/workspace"
    },
    "instances": {
      "personal": {
        "model": "anthropic/claude-opus-4-6",
        "workspace": "~/.openclaw/workspaces/personal",
        "thinkingLevel": "high"
      },
      "work": {
        "model": "anthropic/claude-sonnet-4",
        "workspace": "~/.openclaw/workspaces/work",
        "skills": ["github", "jira", "confluence"]
      },
      "assistant": {
        "model": "anthropic/claude-haiku-3.5",
        "workspace": "~/.openclaw/workspaces/assistant"
      }
    }
  }
}

Agent 配置项

配置项 说明 示例
model 使用的模型 anthropic/claude-opus-4-6
workspace 工作目录 ~/.openclaw/workspaces/personal
thinkingLevel 思考级别 high, medium, low
skills 启用的技能 ["gmail", "calendar"]
sandbox 沙箱配置 { "mode": "non-main" }
contextWindow 上下文窗口 200000

Workspace 结构

~/.openclaw/
├── workspace/              # 默认 workspace
│   ├── AGENTS.md
│   ├── SOUL.md
│   └── skills/
├── workspaces/
│   ├── personal/          # 个人 Agent
│   │   ├── AGENTS.md
│   │   ├── SOUL.md
│   │   └── USER.md
│   ├── work/              # 工作 Agent
│   │   ├── AGENTS.md
│   │   └── skills/
│   └── assistant/         # 助手 Agent
│       └── AGENTS.md
└── openclaw.json          # 主配置

消息路由规则

路由配置

{
  "routing": {
    "rules": [
      {
        "match": {
          "channel": "telegram",
          "sender": "123456789"
        },
        "agent": "personal"
      },
      {
        "match": {
          "channel": "discord",
          "guild": "987654321"
        },
        "agent": "work"
      },
      {
        "match": {
          "session": "main"
        },
        "agent": "personal"
      }
    ],
    "default": "personal"
  }
}

路由规则详解

匹配字段 说明 示例
channel 渠道类型 telegram, discord, slack
sender 发送者 ID 123456789
guild Discord 服务器 ID 987654321
group 群组 ID -100123456789
session 会话 ID main, telegram:user:123
pattern 正则匹配 ^work-

复杂路由示例

{
  "routing": {
    "rules": [
      {
        "name": "个人 Telegram",
        "match": {
          "channel": "telegram",
          "sender": ["123456789", "987654321"]
        },
        "agent": "personal"
      },
      {
        "name": "工作 Discord",
        "match": {
          "channel": "discord",
          "guild": "111222333"
        },
        "agent": "work"
      },
      {
        "name": "工作群组",
        "match": {
          "channel": "telegram",
          "group": "-100111222333"
        },
        "agent": "work"
      },
      {
        "name": "测试环境",
        "match": {
          "pattern": "^test-"
        },
        "agent": "assistant",
        "config": {
          "sandbox": { "mode": "always" }
        }
      }
    ],
    "default": "personal"
  }
}

路由优先级

flowchart TB
    MSG[消息到达] --> R1{匹配规则 1?}
    R1 -->|是| A1[Agent 1]
    R1 -->|否| R2{匹配规则 2?}
    R2 -->|是| A2[Agent 2]
    R2 -->|否| R3{匹配规则 3?}
    R3 -->|是| A3[Agent 3]
    R3 -->|否| D[默认 Agent]

路由按顺序匹配，第一个匹配的规则生效。

会话隔离

隔离级别

级别 说明 使用场景
无隔离 共享上下文 单用户、单用途
渠道隔离 按渠道独立上下文 多渠道接入
用户隔离 按用户独立上下文 多用户共享
会话隔离 每个会话独立 完全隔离
Agent 隔离 每个 Agent 独立 多 Agent 部署

配置会话隔离

{
  "sessions": {
    "isolation": "user",
    "persistence": true,
    "retention": {
      "maxAge": 604800000,
      "maxSessions": 100
    }
  }
}

隔离配置详解

{
  "sessions": {
    "isolation": "user",
    "storage": {
      "type": "file",
      "path": "~/.openclaw/sessions"
    },
    "encryption": {
      "enabled": true,
      "algorithm": "aes-256-gcm"
    },
    "retention": {
      "maxAge": 604800000,
      "maxSessions": 100,
      "cleanupInterval": 86400000
    }
  }
}
配置项 说明 默认值
isolation 隔离级别 user
storage.type 存储类型 file
encryption.enabled 启用加密 true
retention.maxAge 最大保留时间 7 天
retention.maxSessions 最大会话数 100

团队协作场景

场景一：小型团队共享

{
  "agents": {
    "instances": {
      "team": {
        "model": "anthropic/claude-sonnet-4",
        "workspace": "~/.openclaw/workspaces/team",
        "skills": ["github", "jira", "slack"]
      }
    }
  },
  "routing": {
    "rules": [
      {
        "match": { "channel": "slack" },
        "agent": "team"
      },
      {
        "match": { "channel": "discord" },
        "agent": "team"
      }
    ],
    "default": "team"
  }
}

场景二：部门隔离

{
  "agents": {
    "instances": {
      "engineering": {
        "model": "anthropic/claude-opus-4-6",
        "workspace": "~/.openclaw/workspaces/engineering",
        "skills": ["github", "jira", "pagerduty"],
        "sandbox": { "mode": "off" }
      },
      "marketing": {
        "model": "anthropic/claude-sonnet-4",
        "workspace": "~/.openclaw/workspaces/marketing",
        "skills": ["mailchimp", "analytics"],
        "sandbox": { "mode": "non-main" }
      },
      "support": {
        "model": "anthropic/claude-haiku-3.5",
        "workspace": "~/.openclaw/workspaces/support",
        "skills": ["zendesk", "intercom"],
        "sandbox": { "mode": "always" }
      }
    }
  },
  "routing": {
    "rules": [
      {
        "name": "Engineering Discord",
        "match": { "channel": "discord", "guild": "111111111" },
        "agent": "engineering"
      },
      {
        "name": "Marketing Slack",
        "match": { "channel": "slack", "team": "T22222222" },
        "agent": "marketing"
      },
      {
        "name": "Support Telegram",
        "match": { "channel": "telegram", "group": "-100333333333" },
        "agent": "support"
      }
    ],
    "default": "support"
  }
}

场景三：多租户 SaaS

{
  "agents": {
    "dynamic": true,
    "template": {
      "model": "anthropic/claude-sonnet-4",
      "workspace": "~/.openclaw/workspaces/tenants/{{tenantId}}",
      "skills": ["base"],
      "sandbox": { "mode": "always" }
    },
    "limits": {
      "maxAgents": 100,
      "maxSessionsPerAgent": 50,
      "maxTokensPerMonth": 1000000
    }
  },
  "routing": {
    "rules": [
      {
        "name": "Tenant Routing",
        "match": { "header": "X-Tenant-Id" },
        "agent": "tenant_{{tenantId}}",
        "createIfNotExists": true
      }
    ]
  }
}

权限与安全

Agent 权限控制

{
  "agents": {
    "instances": {
      "trusted": {
        "model": "anthropic/claude-opus-4-6",
        "sandbox": { "mode": "off" },
        "permissions": {
          "filesystem": { "read": true, "write": true },
          "network": { "outbound": true },
          "process": { "execute": true }
        }
      },
      "restricted": {
        "model": "anthropic/claude-haiku-3.5",
        "sandbox": { "mode": "always" },
        "permissions": {
          "filesystem": { 
            "read": ["~/allowed/"],
            "write": ["~/allowed/"]
          },
          "network": { 
            "outbound": ["api.example.com"]
          },
          "process": { "execute": false }
        }
      }
    }
  }
}

沙箱模式

模式 说明 适用场景
off 无沙箱 可信 Agent
non-main 非主会话沙箱 团队共享
always 始终沙箱 不信任环境

沙箱配置

{
  "sandbox": {
    "mode": "non-main",
    "container": {
      "image": "openclaw/sandbox:latest",
      "timeout": 30000,
      "memory": "512m",
      "cpu": "0.5"
    },
    "allowlist": ["read", "write", "edit", "bash"],
    "denylist": ["browser", "canvas", "nodes"],
    "networkPolicy": {
      "allow": ["api.example.com"],
      "deny": ["*"]
    }
  }
}

监控与管理

Agent 状态监控

# 列出所有 Agent 实例
openclaw agents list

# 输出示例：
# Name        Model                  Sessions  Status   Last Active
# personal    claude-opus-4-6        3         active   2 min ago
# work        claude-sonnet-4        5         active   5 min ago
# assistant   claude-haiku-3.5       2         idle     1 hour ago

# 查看 Agent 详情
openclaw agents show personal

# 查看路由规则
openclaw routing list

会话管理

# 列出所有会话
openclaw sessions list --agent personal

# 输出示例：
# Session ID              Channel    Sender      Created      Messages
# telegram:user:123       telegram   123456789   2026-02-26   42
# main                    cli        -           2026-02-25   128

# 查看会话详情
openclaw sessions show telegram:user:123

# 导出会话
openclaw sessions export telegram:user:123 --format json

# 删除会话
openclaw sessions delete telegram:user:123

使用量统计

# 查看按 Agent 分组的使用量
openclaw usage --by-agent

# 输出示例：
# Agent       Tokens In    Tokens Out   Cost      Requests
# personal    125,000     45,000      $2.10     128
# work        89,000      32,000      $1.25     95
# assistant   15,000      8,000       $0.15     42

# 按时间范围统计
openclaw usage --by-agent --from 2026-02-01 --to 2026-02-28

最佳实践

1. 合理划分 Agent

划分维度 说明 示例
用途 按功能用途划分 coding, research, assistant
项目 按项目划分 project-alpha, project-beta
团队 按团队划分 engineering, marketing
模型 按模型划分 opus, sonnet, haiku

2. 配置继承

{
  "agents": {
    "defaults": {
      "model": "anthropic/claude-sonnet-4",
      "workspace": "~/.openclaw/workspace",
      "thinkingLevel": "medium",
      "sandbox": { "mode": "non-main" }
    },
    "instances": {
      "opus": {
        "model": "anthropic/claude-opus-4-6",
        "thinkingLevel": "high"
        // 继承 defaults 的 workspace 和 sandbox
      },
      "haiku": {
        "model": "anthropic/claude-haiku-3.5",
        "thinkingLevel": "minimal"
        // 继承 defaults 的其他配置
      }
    }
  }
}

3. 权限最小化

{
  "agents": {
    "instances": {
      "restricted": {
        "model": "anthropic/claude-haiku-3.5",
        "sandbox": { "mode": "always" },
        "permissions": {
          "filesystem": {
            "read": ["~/allowed/read"],
            "write": ["~/allowed/write"]
          },
          "network": {
            "outbound": []
          },
          "process": {
            "execute": false
          }
        }
      }
    }
  }
}

4. 监控和告警

{
  "monitoring": {
    "alerts": [
      {
        "condition": "usage > 1000000",
        "action": "notify"
      },
      {
        "condition": "error_rate > 0.1",
        "action": "notify"
      },
      {
        "condition": "response_time > 10000",
        "action": "notify"
      }
    ],
    "webhook": "https://your-monitoring.com/webhook"
  }
}

故障排查

路由不生效

# 查看当前路由规则
openclaw routing list

# 测试路由
openclaw routing test --channel telegram --sender 123456789

# 输出示例：
# Matched Rule: 个人 Telegram
# Target Agent: personal
# Route Path: default -> rule 1 -> personal

会话隔离问题

# 检查会话隔离配置
openclaw config get sessions.isolation

# 检查会话列表
openclaw sessions list --all

# 重置会话隔离
openclaw sessions reset --agent personal

权限问题

# 检查权限配置
openclaw agents permissions show personal

# 测试权限
openclaw agents test-permission personal --action filesystem.read --path ~/test

# 输出示例：
# Permission: filesystem.read
# Path: ~/test
# Result: ALLOWED

小结

多 Agent 架构是 OpenClaw 的高级功能，为团队和企业场景提供了强大的支持：

  • 灵活路由：按渠道、用户、项目分配不同 Agent
  • 会话隔离：保护隐私，避免上下文混乱
  • 权限控制：精细化的权限管理
  • 成本优化：按用途选择不同模型
  • 可扩展性：支持动态创建和管理 Agent

本篇小结：

  • 理解了多 Agent 架构的价值和场景
  • 掌握了 Agent 配置和消息路由规则
  • 学会了会话隔离的配置方法
  • 了解了团队协作场景的最佳实践
  • 掌握了权限控制和监控管理

更新记录：

  • 2026-02-26：初版发布，基于 OpenClaw v2026.2

系列导航：