# overture-mcp [Health: Active]

**Category:** 🤖 Coding Agents  
**Repository:** https://github.com/SixHq/Overture  
**GitHub Stars:** 643  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/overture-mcp

## Description
MCP Server for Claude Code, Cursor, Cline, Copilot, Github Copilot, Windsurf - Visual AI Agent Plan Execution, Approval Workflow, Plan Visualization, Agent Orchestration. See what your AI is thinking before it writes code. Works with Claude, GPT, Gemini,

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

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

## Documentation

## What overture-mcp does

overture-mcp places a review step between an AI coding agent’s planning and implementation phases. The agent produces a structured plan, and Overture presents it as an interactive flowchart before execution begins. You can inspect individual steps, review expected outputs and risks, compare branch options, and approve the plan when it is ready.

The workflow is intended for tasks where an incorrect interpretation could lead to substantial rework. It supports coding agents including Claude Code, Cursor, Cline, GitHub Copilot, and Windsurf, with installation through each client’s MCP configuration.

## How it works

After you give the coding agent a task, Overture receives the generated plan and renders its nodes as they arrive. The graph can be panned, zoomed, and navigated by selecting nodes. Automatic layout positions the plan, while visual states distinguish pending, active, completed, and failed work. Complexity indicators and a next-node marker provide additional status information.

Before execution, you can enrich the plan by attaching files, entering values, supplying secrets through masked fields, or selecting among alternative branches. Approval starts execution; the interface then shows progress in real time. During the run, you can pause and resume the workflow, rerun nodes, or modify the plan. Nodes can also be inserted between existing steps through the graph’s edges.

## Setup and configuration

The server can be run directly with `npx` and configured as a local stdio MCP server. For Claude Code, use:

```bash
claude mcp add overture-mcp -- npx overture-mcp
```

Cursor, Cline, GitHub Copilot, and Sixth AI use JSON MCP configuration with `npx` as the command and `overture-mcp` as its argument. A global installation is also available:

```bash
npm install -g overture-mcp
```

When the server is working, Overture opens its interface at `http://localhost:3031` after an agent receives a task.

## Tools and capabilities

- Render agent plans as interactive flowcharts with pan, zoom, drag, and click-through navigation.
- Stream plan nodes into the canvas while the agent generates them.
- Display task complexity, dependencies, branches, risks, expected outputs, and execution state.
- Collect string, number, boolean, select, secret, and file inputs at individual nodes.
- Attach files, API keys, and instructions to plan steps.
- Compare the pros and cons of alternative approaches.
- Approve, pause, resume, rerun, or modify execution from the visual workflow.
- Browse and attach MCP tools per node through the MCP marketplace feature.

## Limitations and notes

Overture’s workflow depends on the coding agent producing a plan that the server can render. Execution does not begin until the plan is approved in the interface. The local web interface uses port 3031, so another service occupying that port may prevent it from opening as expected. GitHub Copilot requires the VS Code-specific configuration format rather than `mcpServers`.

_Full upstream README: https://allmcps.com/mcp/overture-mcp/readme_