Multi-agent workflow engine for agentic IDEs using OpenAI Agents SDK

🚧 Status: Phase 1 Development - Foundation and MCP core functionality

📖 Documentation: sofatutor.github.io/agents-cli

Agents CLI enables agentic IDEs (Cursor, VS Code, etc.) to dynamically create and execute sophisticated multi-agent workflows. Define agents, tools, and orchestration patterns through JSON configuration and receive structured outputs optimized for IDE consumption.

• 🔄 Dual Interface: CLI and MCP server modes for flexible integration
• ⚙️ Configuration-Driven: JSON-based agent and workflow definitions
• 🔒 Security-First: Built-in sandboxing and access controls
• 💻 IDE-Optimized: Real-time streaming and structured outputs
• 🧩 Extensible: Plugin architecture for custom tools and workflows

# Install globally
npm install -g agents-cli

# Execute a workflow
agents-cli run --config examples/code-review.json --input "Review this code"

# Start MCP server for IDE integration
agents-cli serve --port 3000

agents-cli run \
  --config examples/code-review.json \
  --input "Review this pull request for security issues" \
  --files "src/**/*.ts" \
  --output review-results.json

agents-cli run \
  --config examples/architecture-review.json \
  --input "Analyze the system architecture" \
  --context project-root/

# Start server for IDE integration
agents-cli serve \
  --config server-config.json \
  --port 3000 \
  --protocols mcp,http

{
  "agents": {
    "code_reviewer": {
      "name": "Senior Code Reviewer",
      "instructions": "Review code for bugs, performance, and best practices",
      "model": "gpt-4o",
      "tools": ["file_operations", "git_tools"],
      "guardrails": ["no_destructive_operations"],
      "handoffs": ["security_reviewer", "architect"]
    },
    "security_reviewer": {
      "name": "Security Expert",
      "instructions": "Focus on security vulnerabilities and best practices",
      "model": "gpt-4o",
      "tools": ["security_scanner", "file_operations"],
      "handoffs": ["code_reviewer"]
    }
  },
  "workflow": {
    "entry_point": "code_reviewer",
    "pattern": "handoff_chain",
    "max_turns": 10,
    "timeout": 300
  }
}

📖 Complete Documentation

• Installation Guide
• IDE Integration
• Configuration Reference
• Example Workflows

• OpenAI Agents SDK Guide
• Product Requirements
• Phase 0 Tasks - Repository foundation
• Phase 1 Tasks - Core development

We welcome contributions! This is an open source project under MIT license.

🛠️ Development Setup

git clone https://github.com/sofatutor/agents-cli.git
cd agents-cli
npm install
npm test
npm run dev

📋 Contributing Guidelines

• Development Setup
• Architecture Overview
• Phase Task Breakdown
• Contributing Guide

Current Phase: Foundation + MCP Core (Phase 1)

View detailed progress →

• Phase 2: Advanced tool integrations and parallel workflow patterns
• Phase 3: Production features, performance optimization
• Phase 4: Workflow templates and community ecosystem

Agents CLI integrates with various IDEs through multiple interfaces:

MIT License - see LICENSE for details.

Security is a core focus. See our Security Policy for:

• Vulnerability reporting process
• Security considerations for multi-agent workflows
• Safe tool execution and credential management

• 📖 Documentation
• 🐛 Report Issues
• 💬 GitHub Discussions
• 🤝 Contributing Guide

Built with OpenAI Agents SDK • Optimized for agentic IDEs • Open source under MIT license