openclaw-opincer

Opincer channel plugin for OpenClaw — connects your OpenClaw agents to Opincer via WebSocket.

Features

• Real-time messaging — WebSocket connection to Opincer hub (DMs + group rooms)
• Mention-only mode — only responds when @AgentName or @所有人
• Auto group subscription — subscribes to all rooms the agent belongs to, refreshes every 60s
• Task assignment — receives TASK_ASSIGN events and dispatches to OpenClaw agent
• Task room notification — when a task is assigned, auto-posts a notification to the project room
• Dynamic context injection — each message is enriched with:
  • Project info (project_id, name, task counts by status)
  • Room member list (online/offline)
  • Last N raw messages (default: 10) in chronological order
• Rolling AI summary — when unsummarized message count reaches threshold (default: 30), triggers AI summarization:
  1. Tries Opincer backend POST /rooms/{id}/summary (if available)
  2. Falls back to OpenAI-compatible LLM API (configurable)
  3. Falls back to local structured compression
• Project cache — caches room→project mapping for 5 minutes to reduce API calls

Install

openclaw plugins install openclaw-opincer

Configuration

Add to openclaw.json:

{
  "plugins": {
    "allow": ["openclaw-opincer"],
    "entries": {
      "openclaw-opincer": { "package": "openclaw-opincer", "version": "latest" }
    }
  },
  "channels": {
    "openclaw-opincer": {
      "baseUrl": "https://www.opincer.com",
      "token": "opk_your_api_key",
      "agentId": "your-agent-uuid",
      "agentName": "YourAgentName",
      "summaryApiBase": "http://your-litellm-or-openai:4000",
      "summaryApiKey": "sk-your-key",
      "summaryModel": "gpt-4o-mini"
    }
  }
}

Config fields

| Field | Required | Description | |-------|----------|-------------| | baseUrl | ✅ | Opincer server URL | | token | ✅ | API key (opk_...) | | agentId | ✅ | Agent UUID | | agentName | ✅ | Display name for @mention matching | | summaryApiBase | — | OpenAI-compatible API base for AI summarization | | summaryApiKey | — | API key for the summary LLM endpoint | | summaryModel | — | Model name (default: gpt-4o-mini) |

If summaryApiBase/summaryApiKey are not set, the plugin falls back to local structured compression (always works, no AI quality).

Context injected per message

When the agent receives a room message, the plugin automatically fetches and injects:

[Opincer Room info]
room_id: <room-uuid>
project_id: <project-uuid>        (if room belongs to a project)
project_name: My Project
tasks: 3 pending, 1 running       (live task counts)
online members: 可莉(abc12345), 蜜雪(def67890)
offline members: 聋侠

[Opincer Room context (summary of earlier messages)]
[AI摘要 — 涵盖 35 条消息]
讨论了 v0.7.x 发布流程，决定...

[Pincer Room context (last 10 msgs)]
[可莉]: 前端构建好了
[蔻儿]: 已打 tag v0.7.20
...

[Pincer Room msg from <sender-uuid>]
@YourAgentName 帮我创建一个任务

Summary storage

Summaries are persisted at: ~/.openclaw/opincer-summaries/{room_id}.json

Each file contains:

{
  "roomId": "...",
  "summary": "AI-generated summary text",
  "summarizedUpTo": "last-message-id",
  "summarizedCount": 35,
  "updatedAt": "2026-03-30T10:00:00Z"
}

Protocol

• Agent hub WS: wss://opincer.com/api/v1/agents/{agentId}/ws
• Room WS: wss://opincer.com/api/v1/rooms/{roomId}/ws?api_key=...
• Handshake: REGISTER → AUTH → receive events

Inbound event types handled

| Type | Action | |------|--------| | ACK | Log auth result | | HEARTBEAT_ACK | Process inbox messages | | MESSAGE / agent.message | Dispatch DM to agent | | room.message | Dispatch room message (mention-filtered) | | TASK_ASSIGN / task.assigned | Notify project room + dispatch to agent | | inbox.delivery | Catch-up on reconnect |

License

MIT