Derpyu520/qq-bridge

Bridge between QQ (SnowLuma OneBot v11) and DeepSeek Harness agents: social simulation, safe MCP tools, slang learning and more.

What it solves

This project bridges QQ messaging with the DeepSeek Harness (DSH) agent framework, allowing users to interact with an AI agent via QQ private messages or group chats. It transforms QQ messages into DSH sessions and routes the agent's responses, questions, and tool approvals back to the QQ user.

How it works

The bridge acts as a middleware between SnowLuma (a OneBot v11 WebSocket client) and the DSH Web API. It maps each QQ chat (private or group) to a unique DSH session within a dedicated "QQ Chat" workspace.

Key technical components include:

  • Communication: Uses WebSocket for QQ connectivity via SnowLuma and a REST/Event-stream API for DSH.
  • MCP Integration: Integrates Model Context Protocol (MCP) servers to give the agent safe capabilities, such as sending messages, checking group status, and performing web searches with SSRF protection.
  • Operational Modes: Supports multiple modes including chat (standard safe chat), closed-agent (full tool access for the owner), and reserved (simulated group member behavior with state machines for activity and silence).
  • Interaction Logic: Handles complex QQ interactions like message citations (replies), slash commands (e.g., /reset), and multi-part message splitting for simulated personas.

Who it’s for

  • Users of DeepSeek Harness who want to bring their AI agents into QQ.
  • Developers looking to create "AI simulated group members" that can autonomously observe and participate in chats.
  • Power users who want to control their DSH instance remotely via QQ.

Highlights

  • Multi-mode Operation: Switch between standard chat, owner-only agent control, and autonomous simulation modes.
  • Safe Tooling: Uses a restricted subset of QQ actions via MCP to prevent unauthorized account control.
  • - Session Management: Automatically manages and persists session mappings between QQ IDs and DSH sessions.
  • Interactive Feedback: Forwards agent-requested user questions and tool approval requests directly to QQ for real-time decision making.
  • Built-in Console: Includes a web-based control panel for real-time mode switching and role configuration.

Related

  • Project
    morpheuslord/HackBot

    An agentic AI system for cybersecurity research and defense that uses multi-agent orchestration, MCP, and ACP to analyze security reports and source code.

  • Project
    github/copilot-plugins

    A collection of official and external plugins for GitHub Copilot that extend its capabilities through skills, hooks, and Model Context Protocol (MCP) servers.

  • Project
    xiaojilele-glitch/WhyBuddy

    SlideRule is a product rehearsal engine that transforms a one-sentence idea into a gated, runnable application preview by generating a comprehensive business structure.

  • Project
    Clearailhc/clearai-dsh

    An ontology discovery and exploration platform that uses a disciplined epistemic loop to build a trustworthy, evidence-based knowledge graph for research.