Skip to content

v0.1.0

Latest

Choose a tag to compare

@github-actions github-actions released this 15 Sep 23:34
· 82 commits to main since this release

axscope v0.1.0

First public release of axscope — an agent-driven browser that reads the page
as text, acts by identity, and renders a cursor for the agent and for whoever is
watching.

Highlights

  • Read the screen as text (snap): an accessibility tree with stable refs,
    scroll state, and a list of clickables the tree does not mark.
  • Act by identity (ref, css=, text=, pos=x,y) with real refusal
    diagnostics when a target is disabled, covered or clipped.
  • Converge, don't sleep (wait, waitgone).
  • Batch with script.
  • A rendered cursor, ripple, HUD and target spotlight injected into the page.
  • Three engines: the extension (your own, already logged-in browser), Chrome
    for Testing, and chrome-headless-shell.
  • MCP server over stdio, sharing the same daemon and refs as the CLI.
  • Multi-session isolation through per-session tab groups in extension mode.

Install

From source:

make install                      # → ~/.local/bin/axscope
axscope install --engine all      # Chrome for Testing + chrome-headless-shell

Or grab the prebuilt binary for your platform below. The tarball bundles the
CLI, the cdpprobe diagnostic tool, the browser extension and the banner script.

Scope of this release

  • Linux and macOS (amd64/arm64). Windows is not supported yet.
  • The extension is loaded unpacked — see the README for setup.

See CHANGELOG.md for the full list of changes.