cordlesssteve/autogen-workspace-mcp
3.2
If you are the rightful owner of autogen-workspace-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.
MCP server for AutoGen workspace management and collaborative agent environments.
AutoGen Workspace MCP Server
This MCP server exposes AutoGen workspace capabilities to Claude Code, enabling seamless integration between Claude Code and the AutoGen multi-agent ecosystem.
Features
🤖 Agent Management
- Create Agents: Create AutoGen agents with sophisticated configurations
- Personality Types: Strategist, Architect, Engineer, Creative, Analyst, Facilitator
- Multi-LLM Support: Ollama, OpenAI, Anthropic providers
- RAG Integration: Optional metaMCP-RAG tool access with 77.8% context reduction
🗣️ Multi-Agent Coordination
- Intelligent Discussions: Start multi-agent discussions with structured rounds
- Consensus Voting: Run democratic decision-making processes
- RAG-Enhanced Tools: Agents can access MCP tools via metaMCP-RAG filtering
📊 Workspace Management
- Status Monitoring: Real-time workspace and agent status
- Infrastructure Health: Redis/Kafka connection monitoring
- Performance Metrics: Agent activity and response time tracking
Available Tools
Tool | Description |
---|---|
autogen__create_agent | Create a new AutoGen agent with specified configuration |
autogen__start_discussion | Start multi-agent discussion with RAG tool access |
autogen__run_consensus_vote | Run consensus voting among agents |
autogen__get_workspace_status | Get comprehensive workspace status |
autogen__list_workspaces | List all available workspaces |
autogen__health_check | Check system health and availability |
Installation
cd /home/cordlesssteve/mcp-servers/autogen-workspace-server
npm install
npm run build
Usage
Add to Claude Code MCP Configuration
Add to your MCP configuration (.claude/mcp.json
or global config):
{
"mcpServers": {
"autogen-workspace": {
"command": "node",
"args": ["dist/index.js"],
"cwd": "/home/cordlesssteve/mcp-servers/autogen-workspace-server"
}
}
}
Example Usage in Claude Code
// Create a business analyst agent
await autogen__create_agent({
agentId: 'business-analyst-01',
name: 'Senior Business Analyst',
personality: 'strategist',
model: {
provider: 'ollama',
modelName: 'mistral:7b-instruct'
},
ragConfig: {
enabled: true,
toolCategories: ['FILE', 'DATA'],
contextReduction: true
}
});
// Start a multi-agent discussion
await autogen__start_discussion({
topic: 'Should we implement microservices architecture?',
agentIds: ['business-analyst-01', 'tech-architect-01'],
rounds: 3,
enableRagTools: true
});
// Run consensus vote
await autogen__run_consensus_vote({
proposalId: 'microservices-proposal',
proposal: 'Migrate to microservices architecture within 6 months',
agentIds: ['business-analyst-01', 'tech-architect-01', 'qa-engineer-01']
});
Integration Architecture
Claude Code Session
↓ (MCP Protocol)
AutoGen Workspace MCP Server
↓ (HTTP API / Direct Integration)
AutoGen Workspace (TypeScript)
↓ (metaMCP-RAG Client)
metaMCP-RAG Server
↓ (RAG Filtering)
MCP Tool Ecosystem (54+ tools)
Configuration
Agent Personalities
Personality | Description | Decision Making | Communication Style |
---|---|---|---|
strategist | Business-focused, ROI-driven | Conservative | Formal |
architect | Technical design expert | Balanced | Technical |
engineer | Implementation focused | Aggressive | Casual |
creative | Innovation focused | Aggressive | Creative |
analyst | Data-driven | Conservative | Analytical |
facilitator | Team coordination | Collaborative | Casual |
RAG Configuration
ragConfig: {
enabled: true, // Enable metaMCP-RAG integration
toolCategories: [ // Filter by tool categories
'FILE', // File operations
'GIT', // Git operations
'SECURITY', // Security scanning
'DATA', // Data processing
'SYSTEM' // System information
],
contextReduction: true // Enable 77.8% context reduction
}
Development
# Development mode
npm run dev
# Build
npm run build
# Watch mode
npm run watch
Connection Methods
1. HTTP API Integration (Preferred)
- Connects to AutoGen workspace via HTTP API
- Requires AutoGen workspace server running on port 3000
- Full feature support
2. Direct TypeScript Integration (Fallback)
- Direct imports from AutoGen TypeScript codebase
- Works without HTTP server
- Limited functionality during development
Error Handling
The server includes comprehensive error handling:
- Validation Errors: Zod schema validation for all inputs
- Connection Errors: Graceful fallback to direct integration
- Execution Errors: Detailed error messages with context
Troubleshooting
Common Issues
1. Connection Refused
# Check if AutoGen workspace is running
curl http://localhost:3000/health
# If not running, the server will use direct integration fallback
2. Agent Creation Fails
- Verify agent ID is unique
- Check personality type is valid
- Ensure model configuration is correct
3. Discussion Timeout
- Increase timeout for longer discussions
- Check agent availability
- Verify workspace status
Health Check
await autogen__health_check();
Returns:
- Connection status
- Integration mode (HTTP API vs Direct)
- Timestamp and system details