TL;DR
webctl is an open-source command-line tool that controls a persistent browser daemon to automate web interactions for humans and AI agents. It prioritizes local filtering and Unix-style pipelines so callers decide what enters their context, positioning itself as an alternative to MCP-style browser tools.
What happened
A new project named webctl provides a command-line interface to control a Chromium browser via a small daemon process. The tool exposes navigation, observation and interaction commands (navigate, snapshot, click, type, select, screenshot, etc.) and relies on semantic element queries based on ARIA roles to target page elements. Sessions are persistent, with cookies stored to disk and named profiles supported. webctl emphasizes keeping control of what information flows into an agent’s context by offering built-in output filtering (interactive-only snapshots, scoping, limits) and compatibility with standard Unix piping tools (grep, jq, head). The architecture uses a JSON-RPC link between the CLI and daemon, and the codebase points to Chromium + Playwright as the browser layer. The project is available on GitHub under an MIT license and can be installed via pip (requires Python 3.11+).
Why it matters
- Gives callers explicit control over what page data is returned to agents, reducing context bloat compared with server-driven MCP responses.
- Fits into Unix toolchains—output can be filtered, truncated or transformed before being consumed by agents or scripts.
- Persistent sessions and named profiles make long-running agent workflows and human takeover simpler.
- A CLI-focused approach makes reproducing and debugging agent actions straightforward by running the same commands locally.
Key facts
- webctl exposes commands for navigation, snapshots, interactions, waiting, session management, and console log handling.
- Element targeting is done with ARIA-role based queries (e.g., role=button name~="Submit"), intended to be robust against CSS changes.
- Snapshot output can be constrained with flags such as –interactive-only, –limit, –within, and formatted as jsonl for downstream tools.
- The tool uses a daemon that manages the browser; the CLI communicates with it over JSON-RPC/TCP or IPC.
- Sessions persist to disk; webctl supports named profiles and explicit save of cookies.
- Install via pip install webctl; setup downloads a Chromium build (~150MB) and Python 3.11+ is required.
- Console inspection commands include streaming (–follow) and level filtering (e.g., –level error or –count for summaries).
- Agent integration helpers include webctl init which creates agent-specific instructions (CLAUDE.md, GEMINI.md, etc.).
- The repository is hosted on GitHub, licensed under MIT, and recent commits show a version bump to 0.1.2.
What to watch next
- Adoption by major agent frameworks and whether webctl becomes a common alternative to MCP — not confirmed in the source
- Further feature development such as broader browser support or official cloud-friendly modes — not confirmed in the source
Quick glossary
- CLI: Command-line interface: a text-based way to issue commands to software, often scriptable and automatable.
- MCP: Model-Client Protocol (as used by some browser tools): a server-driven approach where the server decides what content is returned to the client.
- ARIA roles: Accessibility roles in HTML that describe the purpose of elements (button, link, textbox) and can be used to locate elements semantically.
- JSON-RPC: A lightweight remote procedure call protocol that uses JSON for request and response payloads; used for communication between processes.
- Accessibility tree: A structured representation of a page's elements and their accessible properties, often used by assistive technologies and automation tools.
Reader FAQ
Is webctl open source?
Yes; the repository is on GitHub and the project is distributed under the MIT license.
How do I install webctl?
Installable via pip (pip install webctl) and requires Python 3.11+. A setup step downloads a Chromium build (~150MB).
Does webctl use Playwright or Chromium?
The architecture indicates it manages Chromium and the repository references Playwright as the browser layer.
Can webctl run headless?
Yes; a start mode labeled 'unattended' runs the browser headless.
Do sessions persist across runs?
Yes; cookies persist to disk and webctl supports named profiles and an explicit save command.
webctl Browser automation for AI agents and humans, built on the command line. webctl start webctl navigate "https://google.com" webctl type 'role=combobox name~="Search"' "best restaurants nearby" –submit webctl snapshot –interactive-only –limit…
Sources
- Show HN: Webctl – Browser automation for agents based on CLI instead of MCP
- abhinav-nigam/agent-browser: Browser automation for AI …
- Command-line Browser Automation Tool Exclusively for AI …
- Browser MCP – Automate your browser using VS Code …
Related posts
- Ask HN: How do you safely give LLMs SSH/DB access — tools, proxies, least privilege
- Fast CLI and MCP Server for Managing Lambda Cloud GPU Instances
- Harmony launches AI notetaker for Discord to record, transcribe, summarize