Platform
Stravue for agents and terminals
Most screenshot tools give a program eyes. Stravue gives it a research assistant. Point it at a site with a question in plain words, and it reads the site, picks the pages that answer the question, and returns clean full-page captures on desktop and phone with the reason each page was chosen, a drafted note on every slide, and a deck.
Three ways in
- The MCP server. Add
@stravue/mcpto Claude, Cursor or any client that speaks the Model Context Protocol. Your assistant gets five tools: map a site, capture it, read a capture back, and export the deck or the notes. - The terminal.
npx stravue capturetakes a site and a question and writes the frames into a folder, with the share link printed at the end. - The API. One call starts a capture, one reads it back, and the exports are plain downloads. Keys are made on your account page and can be revoked at any time.
What it can do
- Map every page a public site lists.
- Pick the pages that answer a question, each with a written reason.
- Capture picked or exact pages full length, on desktop and phone.
- Draft a note and numbered comments on every slide.
- Export a deck, a research bundle, plain text, or a single frame.
- Return a branded share link anyone can open.
- Call a webhook when a capture finishes.
What it will not do
By design, and on purpose: it does not sign in to sites, pass paywalls, cookie walls or human checks, click through flows or submit forms, or reach private addresses. When a site blocks it, it says so in plain words and never invents a page. A capture takes minutes rather than milliseconds, so an agent starts one and polls, or waits for the webhook.
Getting started
Create a key on your account page, then add the MCP server to your client or install the terminal tool. Captures spend the same allowance as the app, so a plan covers both. A capture that returns nothing is refunded automatically.
Pricing is on the pricing page. Questions go to support@stravue.com.