NeedMark

English | 日本語

Point at your UI, leave a comment, and hand it to your usual coding agent. NeedMark saves the comment, target/source context and optional screenshot as local files, so the agent can inspect the evidence and you can check the repair.

MIT licensed. It requires no account, MCP server or specific AI provider. The runtime UI does not require React; use the Vite development plugin or a separate review proxy for an existing HTTP development server, including Next.js. NeedMark is independent of hosted NeedBack.

Website · Documentation · npm

Install in your project

Requires Node.js 22.12 or newer. In your application project:

pnpm add -D needmark   # npm install -D needmark / yarn add -D needmark
pnpm exec needmark init
pnpm dev               # your usual development command

Open the app as usual; the NeedMark toolbar appears at the bottom of the page. Examples use pnpm. With npm run the CLI as npx needmark, with Yarn as yarn needmark.

init shows recommended settings and, with your consent, connects NeedMark to the development server. It never touches application source or production builds:

Project	What init changes
Vite (React, Vue and others)	Adds needmark() to plugins in vite.config.*
Astro	Adds vite: { plugins: [needmark()] } to astro.config.*
Next.js	Starts the dev script through needmark dev --, which keeps your usual port and runs Next.js behind it

It also writes needmark.config.ts and targeted Git ignore rules. NEEDMARK=off pnpm dev starts without NeedMark for you; enabled: false in needmark.config.ts turns it off for everyone; pnpm exec needmark remove undoes the connection. Other servers and configs init cannot read are set up by hand as described in Getting started; all noninteractive flags are documented there too.

Review, hand off, verify

1. Hover to preview the target’s tag, name and available component, then select an element, text, area or multiple elements. Shift+click/drag adds multiple targets; Alt/Option+drag selects part of the text; Ctrl/Cmd+click picks a covered target. Enter finishes selection.
2. Write a comment and save it with DOM capture, current-tab capture or no image. Choose Fix, Discuss or Question so the agent knows whether to change code or answer. Each item has comment.md, context.json and an optional screenshot.png.
3. Copy the filtered open comments as instructions for your coding agent. It needs access to the project files and, when used, the images.
4. Have the agent fix the code and reply with its verification. Check the result yourself before resolving the comment.

Agents that support MCP, such as Claude Code and Codex, can register needmark mcp to read feedback with its screenshot, reply, resolve and wait for new feedback. See Connect an AI agent.

pnpm exec needmark export --detail compact
pnpm exec needmark watch
pnpm exec needmark reply <id> 'Adjusted padding; checks passed.' --author Codex
# After the user checks the result:
pnpm exec needmark resolve <id>

Pins and the list support editing, replies and resolving or reopening comments.

DOM capture needs no sharing permission but has rendering limits. Tab capture asks to share the current tab. Passwords and marked regions are masked in saved images; capture and privacy cover the limits and image-free recovery.

Language and documentation

Set language: 'en' or 'ja' in needmark.config.ts; the UI defaults to Japanese for compatibility. A saved browser preference takes precedence. Changing language keeps comments, replies, captured page text and custom component text as written. UI prompt copies use the selected language; CLI exports use an explicitly configured language and otherwise retain English prose.

• Getting started: configuration, selection, handoff, sharing and migration.
• はじめに: the same guide in Japanese.
• Architecture and operating limits: detailed implementation reference in English.
• Documentation index · Contributing · MIT license