Pith Lab
I read a lot for work — and then can't find the important bit when a client asks. Pith remembers what I read, files it into a wiki with sources and hands it back when I need it.
- For
- People who read for work
- Since
- May 2026
- Data
- Hosted in Germany
- Pricing
- Pay per AI action
Reading is easy. Finding it again isn't.
Articles, reports, newsletters — much of it is useful in the moment and gone two weeks later. Bookmarks don't really help. Pith doesn't just keep the link: it pulls out the key points, connects them to what's already there and records where each statement came from.
I originally developed Pith for my own consulting work, and I use it every day.
Save without detours
Browser extension (⌘⇧S), share from your phone, forward by email or follow RSS feeds — whatever is quickest.
A wiki that grows along
Saved texts turn into topic pages. Every statement links back to where it came from.
Morning report
A short report written from your own feeds — read it, or listen to it as three minutes of audio.
Briefings per client
Each engagement gets its own space; new articles are sorted in automatically. Before a meeting there's a briefing with sources.
Check your draft
Before a draft goes out, Pith checks it against your own sources: what's backed up, what isn't.
In Claude & ChatGPT via MCP
Your reading memory becomes a tool for your AI assistant — answers come with citations.
Numbers you can check
When sources contain figures, Pith pulls them into small charts right on the wiki page. Every value keeps its source, and one click shows the exact sentence it came from.
What it looks like
A short tour through the app (26 s), then a few screenshots to click through.
Roughly how it works
01
Capture
Extension, phone, email, RSS.
02
Understand
Fetch the text, summarise it, recognise topics and connections.
03
Remember
A wiki with sources, searchable down to the passage.
04
Deliver
Morning report, briefings, audio — and MCP for your assistant.
Next.js in front, FastAPI and PostgreSQL behind it, language models for text and speech. Runs on my own k3s cluster at Hetzner in Germany.