Zot Chrome Operator: Let Your Terminal AI Agent Drive the Browser via Side Panel
Zot Chrome Operator bridges your terminal-based coding agent zot with Chrome, letting zot operate browser tabs via a browser_action tool. The project by Patrice Eckhart provides a Chrome side panel for chatting with zot and a local WebSocket bridge that executes browser actions.
Installation
No changes to zot itself required. Run:
zot ext install https://github.com/patriceckhart/zot-chrome-operator node "$HOME/Library/Application Support/zot/extensions/zot-chrome-operator/bin/install-cli.js"
The second command creates a zot-chrome shim at ~/.local/bin/zot-chrome. Ensure that directory is on your PATH (add export PATH="$HOME/.local/bin:$PATH" to your shell profile if needed). The shim is auto-refreshed on next zot start, but running install-cli.js makes it available immediately.
Setup Steps
- Build the Chrome extension:
zot-chrome extprints the path to the unpacked build (e.g.,/Users/you/Library/Application Support/zot/extensions/zot-chrome-operator/dist). Load it in Chrome viachrome://extensionswith Developer mode enabled → Load unpacked. - Start the bridge:
zot-chrome startbuilds the extension if needed and starts the bridge in the background. The bridge runszot rpc --no-sessionand listens onws://localhost:9224. - Environment variables:
ZOT_PROVIDER=anthropic ZOT_MODEL=claude-sonnet-4-5 zot-chrome startorPORT=9225 zot-chrome startto customize.
Commands
zot-chrome start– Start the bridge server in backgroundzot-chrome stop– Stop the bridgezot-chrome status– Check bridge statezot-chrome logs– Tail bridge logszot-chrome ext– Print or build the unpacked extension path
Browser Capabilities
The registered zot tool can: list, create, switch, and close tabs; inspect page context; navigate; click; type into native and rich editors; select options; scroll; extract page text; and wait. No session is persisted — the bridge uses zot --no-session.
The bridge logs all actions, viewable with zot-chrome logs.
📖 Read the full source: HN AI Agents
👀 See Also
Local MCP Memory System with Consolidation for AI Conversations
A developer built an MCP server that provides persistent local memory for AI clients, using Qwen 2.5-7B to consolidate conversations into structured knowledge documents every 6 hours. The system runs entirely on your hardware with semantic dedup, adaptive scoring, and FAISS vector search.
CAL: Open-Source Context Optimization Layer for LLM Agents
CAL (Context Assembly Layer) is a Python library that reduces Claude API token usage by 83% through intelligent context selection and compression. It's available via pip install and MIT licensed.
Colony: A Local-First Coordination Layer That Cuts Multi-Agent Handoff Tokens from 30K to 400
Colony is a local-first coordination substrate that reduces multi-agent handoff costs from ~30,000 tokens to ~400 by replacing context replay with compact observations stored in SQLite.
skillcheck: A linter for SKILL.md files that catches cross-agent compatibility issues
skillcheck is a Python tool that validates SKILL.md files against the agentskills.io specification, with unique features including description quality scoring, warnings about Claude-only fields, and file reference validation that aren't available in existing validators.