webctl

webctl is a local browser automation CLI for developers and AI agents that need to read websites and interact with pages. It runs Chromium through Playwright on macOS, Windows and Linux, and supports container deployment. Its main distinction is control over how much page information reaches an agent's context: you can filter results before the agent reads them.

Page output includes structured metadata such as prices, ratings and authors, plus readable Markdown and accessibility snapshots. Filters can focus on interactive elements or a specific page region, while landmark-aware snapshots reduce navigation, footer and sidebar content. Human-readable and JSON output also fit shell scripts and tools such as jq.

For interaction, webctl can click elements by their text or accessibility role, fill forms, handle dropdowns and capture screenshots. It can return the updated page state alongside an action or batch several actions together. Automatic retries handle overlays, cookie prompts and elements that need scrolling into view. Load detection ignores ongoing media streams and WebSocket connections.

The browser can stay visible for human intervention or run headless for unattended work. Sessions persist across commands, cookies stay on disk, and named profiles keep cookies separate. Agent integrations include Claude Code, Codex, Cursor and Gemini CLI. Domain policies restrict the sites the browser can visit, and operating-system credential checks prevent other users from controlling your browser session.