MarkuprPlus MCP
MarkuprPlus records your screen and voice and turns every circled defect into a report with a screenshot for a coding agent.
- 85 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Recording captures the screen and microphone, and pushing findings to GitHub and Linear uses your tokens.
Install
- Works in
Manual install
npm install -g markuprplusFrom the MCP Server section of the README; then `markuprplus integrate claude-desktop` or `codex` and a client restart.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Recording captures the screen and microphone, and pushing findings to GitHub and Linear uses your tokens.
This is third-party code. Review the repository files before installing.
What it does
MarkuprPlus is a Windows and macOS desktop app, a `markuprplus` command line tool and an MCP server. The user picks a window, narrates what is wrong and circles the problem areas on the live screen. Speech is transcribed on the device and the words are aligned to each stroke. The result is a report with one finding per mark, each with its own annotated frame and comment. The MCP server gives an agent `capture_screenshot`, `capture_with_voice`, `analyze_video` and `push_to_github`, and the `analyze` command processes an existing recording.
Who it is for. Frontend developers and testers tired of describing interface bugs in words and attaching screenshots to agent tasks by hand.
Good fit when
- Layout broke on a phone screen and the agent must be shown exactly where.
- Reviewing a finished interface: five remarks in one recording instead of five messages.
- The agent has to see the current screen itself through `capture_screenshot`.
- Findings from a recording must become GitHub or Linear issues.
Not a fit when
- Defects are already described in text with exact paths and code lines.
- Screen and microphone recording is not allowed on the machine.
- You need Linux: the app ships for Windows and macOS.
Example request
Look at my screen and tell me why the sidebar overlaps the content at mobile width, then suggest a CSS fix.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
The app runs on Windows and macOS, and the command line part needs Node.js 20.9 or newer and `ffmpeg` on `PATH`. Recording captures everything visible in the chosen window plus microphone audio, so close windows with private data first. Transcription is local, but pushing findings to GitHub and Linear needs your tokens. The project is young, at version 3.2.0. There is no automatic check of what ended up in the frame: the user is responsible for the recording content.
How to disable. Remove the `MarkuprPlus` entry from the MCP client configuration or run `npm uninstall -g markuprplus`, then uninstall the desktop app. Delete report and recording folders separately.
FAQ
Which MCP tools are there?
Nine: `capture_screenshot`, `capture_with_voice`, `describe_screen`, `start_recording`, `stop_recording`, `analyze_video`, `analyze_screenshot`, `push_to_github` and `push_to_linear`.
How do I connect it to Claude Code?
Add a `MarkuprPlus` server to `~/.claude.json` or `.mcp.json` with command `npx` and args `--yes --package markuprplus markuprplus-mcp`. Cursor and Windsurf use the same shape.
Can an existing recording be processed?
Yes. `markuprplus analyze ./recording.mov` builds a report, `--template github-issue` changes the format and `--no-frames` keeps only the transcript.
Does the transcript go to the cloud?
No, speech is transcribed locally on the device.
Which report templates exist?
`markdown` by default, `json`, `github-issue`, `linear` and `jira`. The desktop app also exports `html` and `pdf`.
Discussion
No questions or reviews yet.
Related
MCP serversA plugin for Claude Code, Codex and Cursor that adds WebMCP tools to your site so browser agents use app logic.
An MCP server shows agent steps on screen: a mechanical spider catches words and branches split work into code, model and you.
An MCP server and a Figma plugin that let an agent read and draw on the canvas safely: with layout checks and no broken screens.
An agent gets the MDMA spec, prompts and validation so it can answer with forms, tables and approvals right inside Markdown
Console, network, screenshots and Lighthouse audits from your open Chrome session to the agent via a DevTools extension
MCP bridge between an agent (Cursor, Claude Code, Codex) and Figma for reading designs and editing them on the canvas
README in short
Three ways to work: a desktop app for recording, a command line tool for finished videos and an MCP server so the agent takes screenshots itself. Reports can be pushed to GitHub Issues and Linear.