# Capture a page as a screenshot, video or PDF as evidence

> Capture a page as a screenshot, video or PDF as evidence: 15 companies fully do it, 5 with a limit, 47 carry the tag. Every listing in Browser automation with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/browser.capture-evidence/
- Markdown: https://www.anchorterminal.com/niches/browser.capture-evidence/index.md (~1,800 tokens)
- Slim: https://www.anchorterminal.com/niches/browser.capture-evidence/index.min.md (~380 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/browser.capture-evidence/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

Four or more companies' docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 15 | 5 | 47 | 4 | 5 |

## What counts as done

A file exists at the path the tool reports, is under the size the channel accepts, and shows the state after the last action.

How agents ask for it: screenshot the result for the PR; record the flow so I can watch it; save this page as a PDF; show me what it looked like when it failed

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **agent-browser (Vercel Inc.)**: screenshot [path] and --full write PNG files, pdf writes a PDF, and record start/stop produces WebM or MP4 when ffmpeg is on PATH. acc-40 built a PR-evidence tool on it; acc-61's video request was closed as completed.
- **Airtop (Switchboard Visual Technologies, Inc. (dba Airtop))**: A screenshot endpoint per window, and session recording with record: true produces one file per window, listed and downloaded through the files API and cleaned up after 7 days.
- **Anchor Browser (Anchor Forge Inc.)**: Recording is on by default and the MP4 is listed and downloaded by session id through the API, kept 7 days on Free and Starter; anchor_* tools include screenshot. acc-35 saw a screenshot call 404 on first try.
- **Browserbase (Browserbase, Inc.)**: Screenshots over CDP write to a local path, every session is recorded by default, and the Recording Downloads API assembles an MP4 per tab by session id, kept 7 days on Free. acc-39 found Stagehand's own context defeated Playwright video recording.
- **Browserless (browserless.io, Inc.)**: screenshot toDisk saves a file and returns a reference, smartscraper has screenshot and pdf formats, screen recording writes WebM on paid plans via /stealth, and session replay with replay=true keeps DOM, console and network for the plan's retention period.
- **BxPlaywright (Ortus Solutions, Corp.)**: pw.screenshot writes a file with fullPage, pdf renders with Chromium, and the CLI's screenshot and pdf verbs and trace.zip give files at named paths.
- **Chrome DevTools MCP (Google LLC)**: take_screenshot writes PNG, JPEG or WebP to filePath with fullPage, and screencast_start writes an MP4 to filePath without ffmpeg since the built-in recorder landed.
- **Hyperbrowser (S2 Labs Inc.)**: web_fetch has a screenshot output format, rrweb web recording is on by default and MP4 video is an option, retrievable for the session.
- **Kane CLI (TestMu AI (formerly LambdaTest))**: Every run writes a sealed .evidence zip at a printed path with screenshot.png per step, annotated copies, HAR and console logs, served locally by kane-cli evidence serve.
- **Kernel (Kernel Technologies, Inc.)**: Replay recordings start, stop, list and download by browser id through the API, and Playwright screenshots write local files.
- **Notte (Notte Labs Inc.)**: browser_screenshot captures the viewport, Playwright writes screenshots to a path, and GET /sessions/{id}/replay returns a presigned MP4 URL valid for an hour.
- **Playwright MCP (Microsoft Corporation)**: browser_take_screenshot writes to --output-dir or a named file, browser_pdf_save comes with --caps pdf, --save-video records and --save-trace keeps a trace. acc-62's screenshot-not-saved report dates from April 2025 and the output-dir rules have since been written down.
- **Puppeteer (Google LLC)**: page.screenshot and ElementHandle.screenshot write PNG files to a path, and page.pdf writes a PDF after fonts load.
- **Selenium (Software Freedom Conservancy)**: The screenshot command returns the viewport as PNG and every binding saves it to a file, element screenshots and Firefox or BiDi full-page captures exist, and the print command produces a PDF with PrintOptions.
- **Skyvern (Ikonomos Inc.)**: Every run keeps a recording and per-step screenshots downloadable from the run page, and a PDF block prints the page as a downloadable artifact.
- **Steel (Nen Labs, Inc.)**: /screenshot and /pdf endpoints and the CLI write files for a URL, sessions record replays viewable in the dashboard, and the Agent Traces ZIP bundles a markdown transcript with screenshots. Retention is 7 days on Launch.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Browser Use (Browser Use, Inc.)**: browserSettings.record: true yields an MP4 in the dashboard after the browser stops, with recordingUrl polled by browser id. Recording is off by default for API runs, never on for ZDR projects, and no screenshot or PDF file endpoint was found.
- **CurlShot (MEDIAPRONET SRL)**: take_screenshot returns a PNG, JPEG, WebP, PDF or scrolling video with a download link good for about a day; files over 6 MB come back as a link. It renders the URL fresh in CurlShot's cloud, so it cannot show a page the agent just acted on.
- **Functionize MCP server (Functionize, Inc.)**: Each run records video and per-step screenshots with side-by-side comparison, but they live in Studio; no tool writes a file to a path the agent names.
- **Lightpanda (Selecy SAS)**: The MCP reference lists a screenshot tool, but the engine has no graphical rendering by the vendor's own description, so what the file shows was not established.
- **TinyFish (Tiny Fish, Inc.)**: video_url holds a signed, expiring recording after a run and the live stream can be captured during it, but there is no screenshot or PDF endpoint; a Browser API session needs the caller's own Playwright for that.

- not this job: Pomerado (Pomerado AI, Inc.): Jobs return typed data, not screenshots, video or PDFs; session recordings are a dashboard setting for the account.

## What stops agents finishing

- a screenshot reported saved that is not on disk (acc-62)
- custom filenames ignoring the output directory
- session replay on an enterprise tier only (acc-35)

## Around it

Before: browser.check-web-app, browser.multi-step-form. After: agent-harnesses.spec-to-build-loop, messaging.send.

## Where this comes from

Weight 4 of 100 in Browser automation: 5 of 61 accounts (acc-08, acc-39, acc-40, acc-61, acc-62); a side product of the testing use case rather than a job on its own

As data: https://www.anchorterminal.com/api/v1/use-cases/browser.capture-evidence.json, or the MCP tool get_use_case with id browser.capture-evidence. Draft, read 2026-11-12 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
