Collaborate on paper reviews with your agents using Plannotator

Introducing Plannotator, the free and open-source tool designed to enhance your paper review process with your agents.

3 min readMachine Learning
Collaborate on paper reviews with your agents using Plannotator
You can use Plannotator to review papers with your agents [P]

Plannotator is a genuinely useful tool for anyone who reads research papers regularly, and the upcoming shared-room feature makes it worth watching now. The ability to annotate PDFs directly in the browser is not new, but coupling that with shareable private annotations and the planned collaborative sessions addresses a specific pain point: how do you keep track of what you think about a paper when you're reading it with others?

The demo shows a clean interface built for the common workflow of reviewing machine learning preprints. You highlight a passage, add a note, and that note stays attached to the text. It sounds trivial, but anyone who has tried to coordinate paper reviews across a lab or study group knows that the typical solution, screenshots, shared docs, Slack threads, breaks down fast. Plannotator keeps the conversation on the page itself. The open-source model matters here; you can inspect the code, modify it, or host your own instance if privacy is a concern. That is a concrete advantage over proprietary tools that bury your annotations in their database.

The room feature, still in development, is what could turn this from a personal utility into a team tool. When it ships, you will be able to invite collaborators into a shared session, annotate together in real time, and leave a persistent record of your group's analysis on each paper. For a lab reviewing a new architecture or a journal club working through a tough submission, that is a shift in how reading happens, less email, more context.

What we like most is that Plannotator does not try to be a platform or a knowledge base. It does one thing: annotate PDFs with agents or people, let you share those annotations, and get out of your way. The creator mentions reviewing DeepSeek v4 as a fun use case, but the real test is whether the room feature feels natural when you have four people in a session trying to parse a figure on page seven. If it does, this tool earns a spot in the weekly workflow. Until then, the private sharing already works, so start using it now. When rooms land, you will already have a library of annotations ready to go.

From Machine Learning

Thought it would be fun to review deepseek v4... sadly not on arxiv yet.

https://plannotator.ai/ is free and open source: https://github.com/backnotprop/plannotator

Read the original at Machine Learning