you can't use this model if it randomly moans between phrases
85 karma · joined October 23, 2019
Screenshot, mouse control (move, click, drag, scroll), and keyboard synthesis (type and press) are all available as CLI commands backed by a native Zig binary
Purpose
- Give the agent a fast, structured overview of files and responsibilities
- Provide jump targets via top-level defs (functions/classes with line numbers)
It uses opentui, the same framework uses by opencode.
It can also render diffs to images, pdf and html. Very useful for agents to share diffs in remote environments like Openclaw or Kimaki
Another benefit is context usage.
The cli can also do a lot more than other MCPs because it uses code snippets run in a stateful sandbox to control the browser, so it can do virtually anything instead of just a few exposed tools like `scroll`, `click`
The big advantage of this architecture is that it that it's very lightweight and starts in less than 200ms. Much faster than the built in Cloudflare tunnel, and no login required
A Raycast porting to the terminal. It will let you run Raycast extensions as TUI apps. Powered by opentui
A browser automation Chrome extension and MCP. It consumes less context than playwright MCP and is more capable: it uses the playwright API directly, the Chrome extension is a CDP protocol proxy via WebSockets.
I use it for automating workflows in development but also filing taxes and other boring tasks
It bridges the CDP protocol from the MCP to the browser, meaning you can do everything Playwright can.
The MCP works using a single tool: execute. It will run Playwright code to control the browser, meaning context usage is small compared to Playwright MCP and the capabilities are more extensive
It will be used to render ANSI inside opencode
Elon Musk considered data centers in space simply for the fact that more solar power is available in space than Earth
I will use it to add support for colored bash tool output in opencode.
---
1. Key differences from Playwright MCP
1. Uses your existing Chrome tabs via an installable Chrome extension instead of launching a new browser window.
2. Exposes only one MCP tool: `execute`, which runs arbitrary Playwright code in a sandbox.
3. The extension communicates with the MCP via Chrome DevTools Protocol (CDP) over WebSocket.
1. Less context bloat: the LLM only needs to reason about a single tool instead of many.
2. More capable: the agent can write any Playwright code to control the browser.
3. Fewer resources: reuses your existing Chrome browser instead of spawning a new instance per MCP project.
4. Human–machine collaboration: you can temporarily disable the extension to:
* Solve CAPTCHAs
* Unstick the LLM
* Intervene manually when needed
1. You can connect to the extension from your own Playwright scripts.
2. This lets you automate tasks in your existing Chrome tabs, instead of running a separate browser instance.
1. The extension does not force-focus Chrome windows during interactions.
2. You can run automation in non-headless mode without Chrome constantly stealing focus and interrupting your work.
1. The extension connects to a localhost WebSocket server started by the MCP.
2. This server proxies MCP commands to the extension.
3. Only tabs where the extension is enabled are accessible to the MCP.
4. Other tabs remain isolated and inaccessible to random agents.
1. The MCP can still cause damage if run unsupervised.
2. You should:
* Run it with permissions enabled, and/or
* Keep it under active human supervision.Each opencode project is a Discord channel.
Start sessions by creating threads.
Supports voice channels via Google Realtime Voice API.
Like Jarvis but for coding agents
Or use urls in your main AGENTS like I do for https://gitchamber.com