Record a repetitive browser workflow once. Replay or schedule it forever. The lean alternative to heavy RPA for legacy ERPs, CRMs, OA systems, intranets, and reporting portals — no code, no cloud, no LLM required.
English · 简体中文 · 日本語 · 한국어 · Español · Français · Deutsch · Português · Русский
Most internal business software — ERPs, CRMs, OA, intranets, reporting portals — has no API. So people become the API: every morning someone opens the same page, picks yesterday's date, applies the same filters, clicks Query, waits for a slow table, clicks Export, and emails the file. Every. Single. Day.
Heavy RPA suites (UiPath, Power Automate) are expensive, need training, and want a desktop install. Pure "AI agents" hallucinate and aren't deterministic. Cloud recorders that demand "read all your data" get rejected by InfoSec.
WebLoop treats the browser DOM as the API. If you can click it, you can automate it.
A lightweight Chrome extension that records your real clicks once and replays them deterministically — on demand or on a schedule — entirely on your machine.
Illustrative UI of the WebLoop side panel (recording a flow, and the per-site permissions panel).
| Feature | What it does | |
|---|---|---|
| ✅ | No-code recording | Capture clicks, inputs, selects, hovers, double-clicks, checkboxes, downloads. |
| ✅ | Deterministic replay | Multi-locator elements (CSS, XPath, stable text, aria-label) + confidence scoring — no hallucinating agents. |
| ✅ | Dynamic date variables | {{today}}, {{yesterday}}, {{month_start}}, {{date:-7}} — reports always query the right window. |
| ✅ | Reliability steps | Wait for text / element / page-stable; auto flow clean-up; static flow audit that flags common pitfalls. |
| ✅ | Scheduling | Manual, daily, workdays, or every N minutes. |
| ✅ | Screenshots & notifications | On success, failure, or when a human is needed (2FA / CAPTCHA / approval). |
| ✅ | Local-first & private | Tasks, logs, settings stay in your browser. Backup/restore as JSON. |
| ✅ | Per-site permissions panel | See, grant, and revoke host access one site at a time. |
| ✅ | Optional AI | Connect OpenAI / Anthropic / Gemini / DeepSeek / Ollama / Groq — off by default, advisory only. |
➡️ See the full Roadmap for what's shipped vs. planned (reliability adapters, resume-from-step, AI repair, team sync).
| WebLoop | Traditional RPA | |
|---|---|---|
| Footprint | Chrome extension | Desktop install / VM |
| Learning curve | Record by clicking | Training & certification |
| Cost | Free & open source (MIT) | Expensive licensing |
| Privacy | Local-first, per-site access | Often cloud / broad access |
| Web compatibility | Built for messy enterprise DOM | Brittle on web apps |
| AI | Optional, deterministic core | Bolted on |
- Record — open your target page, hit Start recording in the side panel, and do the task once.
- Refine — add waits, variable-ize dates, and let the flow audit flag weak spots.
- Test — run it once (or from any step) and read the per-step logs.
- Schedule & relax — pick Daily / Workdays / Interval; WebLoop runs in the background and notifies you.
Full walkthrough: docs/USER_GUIDE.md.
WebLoop's side panel is a React + TypeScript app built with Vite. Build once, then load the generated dist/ folder.
npm install
npm run build # type-checks, builds the side panel, assembles dist/- Open
chrome://extensions/ - Enable Developer mode
- Load unpacked → select the generated
dist/folder
The deterministic core (
service_worker.js,content_script.js) is plain JavaScript, copied intodist/verbatim — only the UI is bundled.
npm run dev # Vite dev server
npm run typecheck # tsc --noEmit
npm run build # production build → dist/
npm run package # build + zip a store-ready archiveWebLoop is local-first and never requests broad "read all your data" access. It asks for one site at a time, only when you record or run there, and every grant is revocable from the in-app Permissions tab. Each baseline permission is explained line-by-line in docs/PERMISSIONS.md.
- Side panel: React 18 + TypeScript, bundled by Vite.
- Automation core: dependency-light, plain-JS Chrome Manifest V3 service worker + content script, copied into
dist/unchanged for auditability. - Storage:
chrome.storage.local— no backend, no telemetry.
Details: docs/ARCHITECTURE.md · Contributing: CONTRIBUTING.md.
What is WebLoop? WebLoop is a free, open-source, no-code browser automation Chrome extension that records and replays repetitive web workflows — form filling, filtering, clicking, file downloads, screenshots, and notifications — as a lightweight, local-first RPA alternative for enterprise and legacy web apps.
Does it need an API or an LLM? No. The core record-and-replay loop is fully deterministic and runs locally without any API or LLM. AI assistance is optional and advisory only.
Is my data sent anywhere? By default, nothing leaves your browser. Only if you explicitly enable the optional AI assistant is a sanitized snippet sent to the provider you configure.
How is it different from UiPath / Power Automate / Automa / Bardeen? WebLoop is intentionally small and focused: local-first, deterministic, with per-site permissions and first-class observability for messy enterprise pages — not a full RPA IDE.
Which sites does it work on?
Any http(s) page you authorize — ERP/OA/CRM systems, intranets, reporting
portals, and other web apps without a usable API.
no-code browser automation · Chrome workflow recorder · lightweight RPA extension · web automation tool · form filling automation · scheduled browser automation · Excel/CSV/PDF download automation · intranet & ERP automation · legacy web app automation · local-first browser agent · record and replay browser tasks.
MIT — built for the modern operator. Powered by simplicity.