markupr

markupR provides an MCP server that enables AI coding agents to interact with your screen capture workflow. When enabled as an MCP server, it can receive structured reports and context-rich markdown generated from your desktop captures, transcripts, and screenshots, making it easier for an agent to understand visual issues and fixes. The server is invoked via npx using the markupr-mcp entry point, so you can run it without a full install and integrate it into your agent workflows. Tools exposed by the MCP server include capture_screenshot (grab the current screen with context hints), capture_with_voice (record screen + mic for a duration and return a structured report), analyze_video (convert existing video into Markdown with extracted frames), analyze_screenshot (process a screenshot through the AI analysis pipeline), start_recording (begin an interactive capture session), and stop_recording (finish the session and run the full pipeline). Use these to build interactive debugging flows where the agent can request specific captures or analyze existing recordings to draft precise markdown reports for your team or CI pipelines.

Prerequisites:

• Node.js installed (includes npm); npx is available with npm 5.2+.
• Internet access to fetch the markupr-mcp package when running via npx.

Installation / setup options:

1. Quick start (no local install):

   • Ensure Node.js is installed.
   • Run the MCP server via npx: npx --yes --package markupr markupr-mcp

2. Local install (optional):

   • Install the package globally (not required for MCP usage, but useful for local development): npm i -g markupr
   • Run the MCP server (if installed globally) using the same entry: npx --yes --package markupr markupr-mcp

3. Integration note:

   • The server is designed to be used as an MCP backend in your agent environment. You can configure your MCP client (e.g., Claude Code or another agent) to point at the server name configured in your MCP settings (e.g., markupR/markupr) and use the provided tools to generate and fetch reports.

Tips and considerations:

• If you’re behind a proxy, ensure npm/yarn proxy settings allow fetching the markupr package from npm registry.
• When using npx, the first run may take extra time to install the package binaries; subsequent runs are faster.
• The MCP workflow assumes the agent can trigger tools like capture_screenshot or analyze_video and then compose the final Markdown reports for downstream use (e.g., PR feedback or issue triage).
• Tools provide structured outputs; prefer using analyze_video for existing recordings and capture_screenshot for on-demand context when debugging issues.
• If you encounter permission or notarization prompts on macOS when using the desktop app, these do not affect the MCP server itself but ensure your environment allows executing the generated reports.