Image, PDF, video, JSON, archives, performance data. One window, one button, every format. Your AI agent's new favorite tool.
brew install --cask scaleassembly/tap/zshot
Click any feature for a closer look.
Screenshot, paginate to PDF, or record a video of any page. Whatever you need, from one place.
Hand the page to an LLM and get back plain text or JSON shaped the way you asked.
An agent clicks, fills forms, and follows your goal before the capture — log in, dismiss the cookie wall, reach the page that matters.
Keep the whole network session: HAR for debugging, WARC for the permanent record.
Record page activity as video, with duration, framerate, and size controls.
Paginate to PDF with a document outline and accessibility tags.
Pull readable page content as clean Markdown — the whole page or one CSS selector.
Save the rendered DOM as HTML or a single-file MHTML archive.
See the page as visitors with protanopia, deuteranopia, tritanopia, or achromatopsia do.
Route the session through HTTP, SOCKS4, or SOCKS5 proxies.
Report chosen CPU cores, memory, GPU, and touch points to the page.
Capture after load, network idle, a selector match, a custom event, or a fixed delay.
Record a DevTools performance timeline with a screenshot filmstrip.
Sample JavaScript CPU and heap activity as pprof profiles.
Anthropic, OpenAI, Gemini, Mistral, DeepSeek, Groq, Hugging Face, Ollama, and local GGUF models.
Run extraction and navigation against models on your own machine.
Serve captures to agents over the Model Context Protocol, locally or remotely.
Run as an HTTP API with auth, concurrency limits, and Swagger docs.
Issue time-limited capture URLs without sharing credentials.
Constrain where the browser may connect, down to CIDR ranges.
Block requests to private address space — the default in server mode.
Every capture brings home an image — and a HAR, a video of what the navigator did, a stderr log, the equivalent CLI command. All of it tucked into the result panel, none of it in your face.