SynapseClean - Local markdown prompt compactor for AI power-users
by•
Free open-source Chrome extension that compacts messy web copy into clean AI prompts. 100% on-device. Optional Gemini Nano. MIT.
Replies
Best
Maker
đź“Ś
Hey Hunters! đź‘‹ Clean prompts shouldn't require sending data to another cloud.
Like many of you, I spend my day copying documentation, GitHub issues, and long technical blog posts to chat panels like Claude, ChatGPT, or Gemini. Every single time, I felt like I was fighting the layout: copying sidebars, cookie banners, tracking scripts, and menus. Not only does this waste precious context tokens (which gets expensive or hits rate limits quickly), but it also dilutes the quality of the LLM's answers.
I wanted a tool that would clean my clipboard instantly, format it as clean Markdown, and—most importantly—never send my copied text to yet another server.
That’s why I built SynapseClean: 1. Zero-Trust Privacy: It runs entirely inside your browser. No database, no telemetry, no subscription fees. 2. Deterministic Pruning: It uses local DOM traversal to remove common boilerplates. 3. Optional Local AI: If you enable the "Semantic engine," it accesses your Chrome browser's built-in Gemini Nano to turn the page text into a highly structured semantic outline before writing it to your clipboard.
It is 100% free and open-source (MIT licensed). I’d love to hear your thoughts, feature requests, and feedback on the local compactor!
It would be great if SynapseClean let you save custom prompt templates per site or domain, so the cleaning rules could match the kind of content you usually grab. Would make it way faster for repeated workflows.
Report
Maker
@derinhrdavh1ay Thanks for the great idea! We are adding Domain-Specific Compaction Profiles to our roadmap so you'll be able to define custom CSS pruning rules and auto-inject prompt templates based on the exact website you're copying from.
Report
Love that it stays fully on-device by default. That choice shows real respect for user privacy while still leaving the door open for Gemini Nano when you need more punch.
Report
Maker
@erenapik Thanks, Eren! Privacy was the core motivation behind building a local-first pipeline. Having Gemini Nano available as an optional semantic engine gives you that cloud-level punch without ever compromising your local environment
Report
love that it's on-device, feels way more private than the alternatives i've tried. one thing that would seal the deal for me is letting us save custom compaction presets for different sites, like a tighter one for arxiv abstracts and a looser one for product pages, so i'm not tweaking the prompt every time.
Report
Maker
@kadir1626034Â Thanks, Kadir! we literally just shipped Domain-Specific Profiles in our latest v1.1.1 update. You can now define custom CSS pruning rules and distinct prompt templates per URL, so everything is fully automated based on the site you're visiting.
Report
Picked a cluttered article and turned it into a usable prompt in one click, which was way smoother than I expected. Love that nothing leaves the browser by default.
Report
Maker
@elmas5mz3 That’s exactly the workflow SynapseClean was built for, Elmas! Keeping everything local means you never have to worry about accidentally exposing sensitive research or data to a third-party cloud backend.
Report
Would love a way to save prompt templates for sites I visit often, like always wrapping cleaned copy from GitHub into a code review prompt. Would save me a bunch of clicks.
Report
Maker
@eliffc46Â Spot on, Elif! We just added this exact functionality (Site Profiles) in the new v1.1.1 update. Youcan now set up a github.com profile that automatically strips out sidebars and injects the extracted code right into your custom review template.
Report
One thing I'd love to see is a per-site memory for prompt templates, so I can have it auto-format HN threads one way and blog posts another without reconfiguring every time. Would make it feel like it actually learns my workflow.
Report
Maker
@recep7qkz I actually just shipped this exact feature in the new v1.1.1 update. You can now set up 'Site Profiles' in the options console, which lets you define custom URL patterns (like Hacker News or specific blogs) and map them to their own distinct prompt templates and CSS pruning rules. It'll automatically apply the right template based on the site you're copying from.
Report
that's a very great one, unfortunately Gemini API is not accessible in my country, i have to use VPN for it, that would just increase my work, but still want to appreciate your work, lovely with great design, cheer up
Replies
Hey Hunters! đź‘‹ Clean prompts shouldn't require sending data to another cloud.
Like many of you, I spend my day copying documentation, GitHub issues, and long technical blog posts to chat panels like Claude, ChatGPT, or Gemini. Every single time, I felt like I was fighting the layout: copying sidebars, cookie banners, tracking scripts, and menus. Not only does this waste precious context tokens (which gets expensive or hits rate limits quickly), but it also dilutes the quality of the LLM's answers.
I wanted a tool that would clean my clipboard instantly, format it as clean Markdown, and—most importantly—never send my copied text to yet another server.
That’s why I built SynapseClean:
1. Zero-Trust Privacy: It runs entirely inside your browser. No database, no telemetry, no subscription fees.
2. Deterministic Pruning: It uses local DOM traversal to remove common boilerplates.
3. Optional Local AI: If you enable the "Semantic engine," it accesses your Chrome browser's built-in Gemini Nano to turn the page text into a highly structured semantic outline before writing it to your clipboard.
It is 100% free and open-source (MIT licensed). I’d love to hear your thoughts, feature requests, and feedback on the local compactor!
GitHub Repository: https://github.com/fujiDevv/synapseclean
Thank you so much for checking it out!
It would be great if SynapseClean let you save custom prompt templates per site or domain, so the cleaning rules could match the kind of content you usually grab. Would make it way faster for repeated workflows.
@derinhrdavh1ay Thanks for the great idea! We are adding Domain-Specific Compaction Profiles to our roadmap so you'll be able to define custom CSS pruning rules and auto-inject prompt templates based on the exact website you're copying from.
Love that it stays fully on-device by default. That choice shows real respect for user privacy while still leaving the door open for Gemini Nano when you need more punch.
@erenapik Thanks, Eren! Privacy was the core motivation behind building a local-first pipeline. Having Gemini Nano available as an optional semantic engine gives you that cloud-level punch without ever compromising your local environment
love that it's on-device, feels way more private than the alternatives i've tried. one thing that would seal the deal for me is letting us save custom compaction presets for different sites, like a tighter one for arxiv abstracts and a looser one for product pages, so i'm not tweaking the prompt every time.
@kadir1626034Â Thanks, Kadir! we literally just shipped Domain-Specific Profiles in our latest v1.1.1 update. You can now define custom CSS pruning rules and distinct prompt templates per URL, so everything is fully automated based on the site you're visiting.
Picked a cluttered article and turned it into a usable prompt in one click, which was way smoother than I expected. Love that nothing leaves the browser by default.
@elmas5mz3 That’s exactly the workflow SynapseClean was built for, Elmas! Keeping everything local means you never have to worry about accidentally exposing sensitive research or data to a third-party cloud backend.
Would love a way to save prompt templates for sites I visit often, like always wrapping cleaned copy from GitHub into a code review prompt. Would save me a bunch of clicks.
@eliffc46Â Spot on, Elif! We just added this exact functionality (Site Profiles) in the new v1.1.1 update. Youcan now set up a github.com profile that automatically strips out sidebars and injects the extracted code right into your custom review template.
One thing I'd love to see is a per-site memory for prompt templates, so I can have it auto-format HN threads one way and blog posts another without reconfiguring every time. Would make it feel like it actually learns my workflow.
@recep7qkz I actually just shipped this exact feature in the new v1.1.1 update. You can now set up 'Site Profiles' in the options console, which lets you define custom URL patterns (like Hacker News or specific blogs) and map them to their own distinct prompt templates and CSS pruning rules. It'll automatically apply the right template based on the site you're copying from.
that's a very great one, unfortunately Gemini API is not accessible in my country, i have to use VPN for it, that would just increase my work, but still want to appreciate your work, lovely with great design, cheer up