Borrowing the browser session is the whole idea, and it is exactly why the permission model has to be strict: everything starts off, the plugin source gets reviewed before you enable it, and permissions reset the moment a plugin updates.
A local CLI plus a Chrome extension. The extension bridges your browser to a local server, and plugins turn a site's own web APIs into tools your agent can call using the session you are already signed in with — not screenshots, not DOM scraping.
- 100+ plugins covering roughly 2,000 tools, including Slack, Discord, GitHub, Jira, Notion, Figma, AWS and Stripe, installed with one command and usable immediately with no restart
- Built-in browser tools that need no plugin at all — screenshots, clicking, typing and network capture, on any tab
- Plugin scaffolding in one command, publishable to npm so anyone can install it; or point your AI at a site and have it discover the APIs and build the plugin for you
- A CLI mode for agents that do not speak MCP
- A permission model that starts closed: nothing executes until you enable it, your AI reviews the plugin source before you do, permissions reset when a plugin updates, and each plugin or tool can be set to Off, Ask or Auto
Node.js 22+ and Chrome. Install with `npm install -g @opentabs-dev/cli` and run `opentabs start`, then load the extension from `~/.opentabs/extension` in `chrome://extensions/` with Developer mode on. Plugins are added with `opentabs plugin install <plugin-name>`. Everything runs locally with a full audit log and no cloud component; anonymous telemetry is on by default and can be opted out of. It is not affiliated with or endorsed by any of the third-party services it reaches.
One command — npm install -g @opentabs-dev/cli