Logos Files

Logos 0.2.1 makes the browser reach places it could not before: controls inside iframes, points with no element behind them, and page state that only JavaScript can read. Updating from 0.2.0 is a drop-in replacement - settings.json, logs, session journals and projects in %LOCALAPPDATA%\Logos are kept.

Browser

• Clicks by coordinates. browser_click keeps its ref argument and now also takes x/y: CSS pixels from the top-left corner of the visible page, the same pixels as browser_screenshot (which is never scaled). This is what canvas, image maps and widgets that only listen to raw mouse events need. A point outside the viewport is rejected with a readable message instead of clicking blindly, and button selects left, right or middle.
• Iframes are part of the page. browser_snapshot now walks every frame, so a control inside a cross-origin iframe (embedded checkout forms, captcha widgets, third-party players) gets a numbered ref like any other element and is marked (frame N: host). Click, fill and hover work on it by that ref - no separate "switch to frame" step. browser_read appends each frame's text as Frame N lines, because cross-origin frames never appear in the top document text.
• browser_frames lists the frames with their index, URL and name.
• browser_eval runs one JavaScript expression or function in the page, or in a frame by index, and returns the value as text (capped at 6 000 characters, wrapped as untrusted page content). Every call asks for your approval and shows the script, because page JavaScript can read cookies and storage and fetch with the page's session.

Verified on webscraper.io/bot-check: the "CDP detection with a click within cross domain iframe and screenX, screenY" test now passes, because the checkbox inside its sandboxed iframe finally has a ref (the click reports a real isTrusted event with screenY 485). The one test still failing there, "Popular Browser", comes from the navigator.userAgentData brand list of a plain Chromium build - not from automation signals. Closing that gap is stealth work (patchright / rebrowser-patches), not done yet.

Download

Both builds keep settings, logs, cache and project data in %LOCALAPPDATA%\Logos.

Notes

• Built for Windows 10/11 (x64). On Linux, run from source - see the README.
• Browser mode drives a real Chromium through Playwright, which is not bundled to keep the download small. On first use Logos offers the download (~320 MB) into a per-user folder and shows per-file progress; Settings → Browser shows the same status and button.
• Vision (Computer Use) needs a provider that supports it - OpenAI, Anthropic or Google with a paid key. OpenRouter and custom endpoints are not used for Vision.
• License: MIT.