One agent for the open web
Checkout, a job application, a multi-step wizard, reading one page to write on another — no integration, no scripts.
OpenSidebar is an open-source Chrome extension that puts an autonomous agent in your side panel — describe a task in plain English and it sees the page, clicks, types, and carries multi-step work across tabs to done.
Bring your own API key. No subscription, no telemetry, no backend — everything runs in your browser.
The Chrome Web Store listing is on the way. Until then, load the unpacked build — it takes one clone and one command.
A signed listing so anyone can add OpenSidebar straight from Chrome.
Coming soon# clone & build git clone https://github.com/krisshkodrani/OpenSidebar cd OpenSidebar corepack enable corepack pnpm install corepack pnpm run dist # then in Chrome chrome://extensions → Developer mode → Load unpacked → select dist/
Requires Node.js 22+ and a supported provider API key.
No mockups, no cherry-picked frames — these are real sessions the agent drove end to end.
Checkout, a job application, a multi-step wizard, reading one page to write on another — no integration, no scripts.
The same agent drives ServiceNow — creating incidents, ordering from the catalog, filtering lists. One agent, every workflow.
Leave it watching a page and it tells you the moment something changes — flagged “back in stock” the instant it flipped.
Paste a key, pick your models, go. Fireworks, OpenRouter, Moonshot, or Xiaomi — nothing hardcoded.
Eleven scenes in three acts — the open web, staying in control, and the ServiceNow finale. Narrated, 3:35.
Recorded live — no scripts, no post-editing. Executor: Kimi K2.7 Code (vision) · Planner: GLM 5.2 · Fireworks AI
A vision model works from the live screenshot and DOM — reads charts, zooms into fine print, and handles pages that defeat text-only bots.
Clicking, typing, scrolling — plus file upload and download, tab and window management, and structured table, chart, and filter extraction.
Four approval modes. Consequential actions pause for you, and form submits are dry-run first — you approve the exact field values before they’re sent.
A planner decomposes the task, an executor drives it, a verifier confirms the result — and stuck runs escalate to a stronger model instead of burning turns.
A local profile you review yourself, per-site skills learned from successful runs, and checkpoints that survive restarts. Sensitive fields are consent-gated and encrypted.
Leave it watching a page and it tells you the moment the page changes — a price, a status, a new listing.
Voice input transcribes straight into the composer — describe the task out loud and let it run.
Every session produces a full-fidelity local trace. Keys live in Chrome storage; traffic goes browser → provider and nowhere else.
Your API key stays in Chrome storage; page context goes only to the provider you configure. No analytics, no tracking, no hosted relay, no OpenSidebar servers.
The repo adds a full observability workspace for developers: replay any session turn by turn — scorecards, tool calls, per-turn screenshots, and token and cost roll-ups. Traces never leave your machine.