# md-feedback

Lets your agent read annotations on a markdown plan: fixes, questions, highlights. Useful when the plan comes before the code.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 5
- Author: yeominux
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/yeominux/md-feedback
- Updated: 2026-03-23
- HTML: https://affpapa.org/en/nevibecode/mcp/md-feedback

## What it does

md-feedback is an MCP server and a VS Code extension. You open a `.md` plan, select some text and press `1`, `2` or `3` to leave a highlight, a fix or a question. The agent picks those annotations up over MCP and starts working. No export step, no copy-paste.

Under the hood there are 27 MCP tools, quality gates that evaluate pass or fail on their own, checkpoints, and session handoffs that carry context forward. Annotations live in the file as HTML comments, so they survive commits and merges. An agent can roll back its last change, and writes to `.env`, `credentials` or `node_modules` are blocked.

MCP needs Node.js 18+ because the server runs through `npx`. Config paths differ per client: `.claude/mcp.json` for Claude, `.cursor/mcp.json` for Cursor. Export-only agents, such as the current Copilot flow, cannot read annotations directly.

## Requirements
- Needs: Node.js (commands run via npx)

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add md-feedback -- npx -y md-feedback
```

### Codex

**Platform CLI**
```bash
codex mcp add md-feedback -- npx -y md-feedback
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "md-feedback": {
            "command": "npx",
            "args": [
                "-y",
                "md-feedback"
            ]
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "md-feedback": {
            "type": "stdio",
            "command": "npx",
            "args": [
                "-y",
                "md-feedback"
            ]
        }
    }
}
```

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add md-feedback npx -y md-feedback
```

## When to use it
- check an agent's plan before it writes any code
- mark the parts of a plan that need rework
- ask the agent a question inside the plan text
- carry context from one agent session to the next
- set conditions the agent must meet before moving on

## Example prompts
- Read the annotations in `plan.md` through md-feedback and apply the fixes
- Answer the questions I left in the plan
- Generate a handoff for the next session

## Similar
- [ce-resolve-pr-feedback](https://affpapa.org/en/nevibecode/skills/ce-resolve-pr-feedback.md) — skill · 3200 installs
- [pr-feedback-quality-gate](https://affpapa.org/en/nevibecode/skills/pr-feedback-quality-gate.md) — skill · 1800 installs
- [code-walk-thru](https://affpapa.org/en/nevibecode/skills/code-walk-thru.md) — skill · ★ 2021
- [pr-address-feedback](https://affpapa.org/en/nevibecode/skills/pr-address-feedback.md) — skill · 2 installs
- [agents](https://affpapa.org/en/nevibecode/rules/agents.md) — rule · ★ 40451
- [docs-review](https://affpapa.org/en/nevibecode/skills/docs-review.md) — skill · 2000 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md