Roundtable MCP
🏛️ 圆桌 MCP (roundtable-mcp)
基于角色的混合多模型编排 MCP 服务器
roundtable-mcp 是一个开源的模型上下文协议(MCP)服务器,它编排一组专业 AI 模型——将本地零成本模型(通过 Ollama)与云端 API(Anthropic Claude、Google Gemini、OpenAI、DeepSeek 等)相结合。
与对通用答案进行投票不同,Roundtable 为每个模型分配不同的专家角色(架构师、实现工程师、叙事专家、系统重构专家、深度推理专家),并将它们的见解综合成统一顾问委员会报告。
✨ 主要特性
基于角色的面板:每个模型从指定的领域视角出发,而非提供通用的重复回答。
混合执行:无缝混合本地模型(Ollama Gemma 4、Qwen Coder、Llama 3)与云端 API(Claude 3.7/4.6、Gemini 3.5、DeepSeek R1、GPT-4o)。
并行扇出:
run_roundtable使用Promise.allSettled()并行查询所有活跃面板模型,实现快速执行和高容错性。配置驱动:随时通过编辑
config.json添加或调整模型,无需修改代码。跨平台 MCP:原生兼容 Claude Code、Cursor、Windsurf、Antigravity、OpenCode 和 VS Code。
Related MCP server: Councly MCP Server
📦 快速开始
1. 安装
git clone https://github.com/YOUR_USERNAME/roundtable-mcp.git
cd roundtable-mcp
npm install
npm run build2. 配置
将 config.example.json 复制为 config.json:
cp config.example.json config.json为云模型设置所需的环境变量:
export ANTHROPIC_API_KEY="sk-ant-..."
export DEEPSEEK_API_KEY="sk-..."
export OPENAI_API_KEY="sk-..."
export GEMINI_API_KEY="AIzaSy..."🛠️ MCP 工具参考
工具名称 | 描述 |
| 在所有启用的 Roundtable 专家上并行运行提示,并输出结构化的顾问委员会报告。如果启用,自动记录到 Obsidian。 |
| 将提示路由到特定专家模型,或根据关键词触发自动检测( |
| 列出所有注册的专家模型、它们的角色以及当前提供商状态。 |
| 列出内置预设角色(QA 测试员、CI/CD 工程师、安全审计员、数据库架构师等)。 |
| 在 Obsidian Vault 中搜索过往 Roundtable 会话笔记,获取相关上下文片段。 |
| 手动或自动将自定义笔记和会话直接记录到 Obsidian Vault。 |
📋 预置角色目录(role_preset)
无需编写自定义角色描述,您可以在 config.json 中通过 "role_preset" 使用内置预设键:
qa_tester:QA 与边界案例测试专家cicd_engineer:DevOps 与 CI/CD 流水线专家security_auditor:网络安全与漏洞审计员database_architect:数据库架构师与查询优化专家architect:首席系统架构师与编排者performance_tuner:底层性能与重构专家frontend_ux:前端与 UI/UX 无障碍专家game_engineer:游戏引擎与机制专家tech_writer:技术文档与 API 编写专家
{
"id": "claude-qa",
"name": "Claude Sonnet 4.6",
"role_preset": "qa_tester",
"provider": "anthropic",
"model": "claude-3-7-sonnet-20250219",
"enabled": true
}🎯 为模型分配专业技能(skill)
是的!您可以根据每个模型的角色,为其分配特定的领域技能指南、编码标准或检查清单。
在 config.json 中使用 "skill"(或 "skill_prompt")属性,将自定义指令注入到该模型的系统提示中:
{
"id": "gemma-godot",
"name": "Gemma 4 (Local)",
"role": "GDScript Implementation Specialist",
"skill": "Follow Godot 4 GDScript standards: strict static typing, signal decoupling, and node path verification.",
"provider": "ollama",
"endpoint": "http://localhost:11434/api/generate",
"model": "gemma4-26b-128k:latest",
"enabled": true
}📓 Obsidian Vault 集成(长期记忆)
roundtable-mcp 直接连接到您的本地 Obsidian Vault,自动记录会话笔记并提供跨会话的持久记忆——即使对于上下文窗口较小的本地模型也是如此!
config.json 中的配置:
{
"obsidian": {
"enabled": true,
"vault_path": "C:/Users/yourname/Documents/ObsidianVault",
"folder": "Roundtable/Sessions",
"auto_save": true,
"max_context_notes": 3
}
}自动记录会话:每个
run_roundtable会话都会格式化为干净的 Markdown 笔记,包含元数据、日期和标签(#roundtable #council #gemma #claude)。记忆检索:使用
search_obsidian_memory获取紧凑的过去笔记片段,使上下文窗口较小的模型也能保持信息同步,避免上下文溢出!
🔧 与 AI 编码平台集成
Claude Code / Claude Desktop / Antigravity / OpenCode
将 roundtable-mcp 添加到您的 MCP 配置文件(mcpServers 部分):
{
"mcpServers": {
"roundtable": {
"command": "node",
"args": ["/path/to/roundtable-mcp/dist/index.js"],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-...",
"DEEPSEEK_API_KEY": "sk-...",
"OPENAI_API_KEY": "sk-...",
"GEMINI_API_KEY": "AIzaSy..."
}
}
}
}⚙️ 向 config.json 添加新模型
您可以轻松添加任何兼容 OpenAI 的提供商(DeepSeek、Groq、Together AI、Mistral、LM Studio、vLLM):
{
"id": "qwen-coder-local",
"name": "Qwen 2.5 Coder (Local)",
"role": "Low-Level Shader & GPU Optimization Specialist",
"provider": "ollama",
"endpoint": "http://localhost:11434/api/generate",
"model": "qwen2.5-coder:32b",
"enabled": true
}💡 完全免费的多模型栈示例
您可以通过结合本地 Ollama 模型和免费云端 API 层级(GitHub Models、Google AI Studio、Groq),完全免费运行一个完整的多模型 Roundtable 设置:
{
"models": [
{
"id": "gpt4o-free",
"name": "GPT-4o (GitHub Free)",
"role": "Systems Performance & Refactoring Specialist",
"provider": "openai-compatible",
"endpoint": "https://models.github.ai/inference/chat/completions",
"model": "gpt-4o",
"api_key_env": "GITHUB_TOKEN",
"enabled": true
},
{
"id": "gemini-free",
"name": "Gemini 2.5 Flash (Google Free)",
"role": "Lead Architect & Synthesizer",
"provider": "google",
"model": "gemini-2.5-flash",
"api_key_env": "GEMINI_API_KEY",
"enabled": true
},
{
"id": "gemma-local",
"name": "Gemma 4 (Local Ollama)",
"role": "Primary Implementation Engineer",
"provider": "ollama",
"endpoint": "http://localhost:11434/api/generate",
"model": "gemma4-26b-128k:latest",
"enabled": true
},
{
"id": "deepseek-free",
"name": "DeepSeek R1 (Groq Free)",
"role": "Deep Reasoning & Math Specialist",
"provider": "openai-compatible",
"endpoint": "https://api.groq.com/openai/v1/chat/completions",
"model": "deepseek-r1-distill-llama-70b",
"api_key_env": "GROQ_API_KEY",
"enabled": true
}
]
}📄 许可证
MIT 许可证。欢迎使用、修改和分发!
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- Flicense-quality-maintenanceOrchestrates multiple AI models (Gemini, OpenAI, Claude, local models) within a single conversation context, enabling collaborative workflows like multi-model code reviews, consensus building, and CLI-to-CLI bridging for specialized tasks.
- AlicenseAqualityDmaintenanceEnables AI assistants to create council hearings where multiple LLMs (Claude, GPT, Gemini, Grok) debate topics and synthesize verdicts with trust scores and diverse perspectives.210Apache 2.0
- Alicense-qualityBmaintenanceEnables multi-strategy AI orchestration including council decision review, debate, brainstorming, evaluation, and spec review, with support for multiple LLM providers and advisor personas.20MIT
- Flicense-qualityCmaintenanceEnables users to leverage a mixture-of-agents architecture where multiple AI models discuss a question in parallel and a president model synthesizes the best answer.17
Related MCP Connectors
Build, validate, and deploy multi-agent AI solutions from any AI environment.
Enterprise AI Control Plane: governance, guardrails, spend tracking, compliance & smart routing.
Deliberation + live 5-model council divergence over the Omnarai multi-AI attributed corpus.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/hardc0l2e/roundtable-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server