← sources

Browser-native automation: pulled a page's table into a Sheet with no server, no cloud, no API key

source: reddit (reddit:selfhosted)published 2026-10-20 21:06ingested 2026-10-21 02:38

retained text

Most automation ends up as "stand up n8n or a scraper on a box and maintain it." For small jobs I wanted the opposite: something that runs locally with nothing to host.

So I built it in the browser: a 4-step workflow that extracts a table off the current page, cleans it, and appends it to a Google Sheet. It runs client-side in the extension. Page parsing and the optional AI steps happen on-device. The only outbound call is the Sheets write, to your own Sheet.

Relevant bits for this crowd:  
\- \*\*No server, no container\*\*: it's a Manifest V3 extension; workflows run in the browser  
\- \*\*Local-first data\*\*: workflow state lives locally; no account needed to start  
\- \*\*On-device AI option\*\*: WebLLM / transformers.js for summarize/classify/extract steps, or point it at a local Ollama; cloud providers are optional  
\- \*\*Survives a closed tab\*\*: scheduled and waiting runs use browser alarms (not a 24/7 headless runner; that's where a real server still wins)  
\- \*\*OpenAPI node\*\* for anything with a spec, so it isn't a walled garden

Trade-offs: it runs when your browser is available, and on-device models are smaller than the cloud giants. For always-on headless jobs I'd still self-host. For "pull this into a Sheet without maintaining infra", it replaced a cron plus script for me.

Disclosure: I build it (AWFlow, free). It's a browser extension, not a self-hosted server app. I'm happy to share the workflow template or answer architecture questions.

Where do you draw the line between "just self-host it" and "a browser extension is enough"?

document fetched 2026-10-21 00:28 · canonical https://reddit.com/r/selfhosted/comments/1xb47hi/browsernative_automation_pulled_a_pages_table