Overture

Project Overview#

Overture is a visual plan execution and approval workflow tool designed for AI coding agents (Claude Code, Cursor, Cline, Copilot). It runs as a local MCP (Model Context Protocol) server, rendering execution plans as interactive flowcharts for user review and approval before the agent begins writing code.

Core Capabilities#

Visual Interaction#

• Interactive Flowcharts: Renders AI agent execution plans as node graphs, clearly showing relationships between steps and branching paths
• Node Details View: Click any node to view complete context and instructions for that step
• Decision Path Selection: When agents propose multiple solutions, users can manually select execution paths

Execution Control#

• Pre-approval Mechanism: Agents are blocked from writing any code until users approve the plan
• Context Attachment: Attach files, documents, API keys, and instructions as context for specific steps
• Real-time Execution Monitoring: Nodes display active, completed, or error status in real-time

System Integration#

• MCP Protocol Support: Runs as an MCP Server, compatible with any MCP-supporting AI coding agent
• Local-first: Runs entirely locally, data stays on localhost
• WebSocket Communication: Supports real-time bidirectional communication

Problems Solved#

• Lack of Transparency: Traditional AI coding agents don't show complete plans before writing code
• High Cost of Misunderstanding: Discovering agent misunderstandings after extensive code has been written
• Text Plan Limitations: Text plans cannot show step relationships, branching paths, and context requirements

Installation#

Quick Start#

# Run directly with npx (no installation required)
npx overture-mcp

Global Installation#

npm install -g overture-mcp

Configuration for Different AI Agents#

Claude Code

claude mcp add overture-mcp -- npx overture-mcp

Cursor Edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "overture": {
      "command": "npx",
      "args": ["overture-mcp"]
    }
  }
}

Cline (VS Code Extension)

{
  "mcpServers": {
    "overture": {
      "command": "npx",
      "args": ["overture-mcp"]
    }
  }
}

GitHub Copilot Create .vscode/mcp.json in project root:

{
  "servers": {
    "overture": {
      "command": "npx",
      "args": ["overture-mcp"]
    }
  }
}

Configuration Options#

Advanced Configuration Example#

{
  "mcpServers": {
    "overture": {
      "command": "npx",
      "args": ["overture-mcp"],
      "env": {
        "OVERTURE_HTTP_PORT": "4000",
        "OVERTURE_WS_PORT": "4001",
        "OVERTURE_AUTO_OPEN": "false"
      }
    }
  }
}

Workflow#

1. Submit a task to the AI agent (e.g., "Build a REST API with authentication")
2. Agent generates a detailed plan, broken into steps with branching paths
3. Overture displays an interactive graphical plan in the browser
4. User reviews and enriches the plan (view node details, attach context, select decision paths)
5. User approves the plan, agent begins execution
6. Watch progress in real-time, node status updates (active, completed, failed)

Project Structure#

Overture/
├── packages/
│   ├── mcp-server/     # MCP server implementation
│   └── ui/             # Web UI interface
├── examples/
│   └── ecommerce-plan.xml  # Sample plan file
├── prompts/            # Prompt-related files
└── assets/             # Resource files

Technical Specifications#

• Primary Languages: TypeScript (82.1%), JavaScript (16.8%), HTML (1.1%)
• Protocol: MCP (Model Context Protocol)
• Communication: WebSocket + HTTP
• License: MIT License
• Developer: SixHq
• Current Version: 0.1.6