ProofShot: CLI for AI Agents to Verify UI Code with Browser Recording
What ProofShot Does
ProofShot is a CLI tool that gives AI coding agents visual verification capabilities. It allows agents to see what the UI they build actually looks like in the browser, detect layout issues, and capture console errors.
How It Works
The tool operates through three main commands:
proofshot start --run "npm run dev" --port 3000- Launches your dev server, opens headless Chromium, and starts recording video- Your AI agent then executes actions like
proofshot exec navigate "http://localhost:3000"andproofshot exec screenshot "homepage"to navigate, click, fill forms, and take screenshots proofshot stop- Collects errors, stops recording, trims dead time, and generates proof artifacts
Output and Features
ProofShot generates a standalone HTML file containing:
- Video playback of the browser session synced with an action timeline
- Screenshots taken during the session
- Element labels for each action
- Browser console errors captured during the session
- Server logs scanned with pattern matching for JavaScript, Python, Go, Rust, and other languages
- PR-ready artifacts including SUMMARY.md and formatted output for pull requests
- Visual diff comparison against baselines
Technical Details
The tool is:
- Built on agent-browser from Vercel Labs (described as "far better and faster than Playwright MCP")
- Not a testing framework - the agent doesn't decide pass/fail, it just provides evidence
- Agent-agnostic - works with Claude Code, Cursor, Codex, Gemini CLI, Windsurf, and any MCP-compatible agent
- Packaged as a skill so AI agents know exactly how it works
- Open source with MIT license
Installation and Setup
$ npm install -g proofshot
$ proofshot install
The tool automatically trims dead time from recordings, so you see only what the agent actually did, not idle waiting periods.
📖 Read the full source: HN LLM Tools
👀 See Also
Unsloth and NVIDIA Collaborate to Speed Up LLM Training by ~25%
Unsloth and NVIDIA release optimizations for LLM training: caching packed-sequence metadata (~14.3% speedup) and double-buffered async gradient checkpointing (~8% speedup), with no accuracy loss. Auto-enabled on RTX laptops, data center GPUs, and DGX Spark.
MCP Server for Italian Train Data: Real-Time Delays, Departures, and Schedules in Claude
A developer built an unofficial MCP server for Trenitalia that provides five tools for querying Italian train data through Claude, including real-time departure/arrival boards, train tracking, and schedules with live delay enrichment.
Depct tool collects runtime data to help Claude debug production issues
Depct is a tool that collects runtime instrumentation from Node.js apps, builds graphs from the data, and feeds it to Claude via AWS Bedrock to help debug intermittent production failures. It also generates architecture diagrams and dependency maps from runtime behavior.
OpenClawDreams: A Dream Simulator Extension for OpenClaw Agents
OpenClawDreams is an extension that adds a background reflection process and nightly dream cycle to OpenClaw agents. It captures encrypted conversation summaries to a local SQLite database, processes them during background cycles, and generates consolidated insights that get pushed into the agent's persistent memory.