Free AI Science Tools
The Tools
Built on One Idea: Bring Your Own AI
Most AI tools for researchers are subscription services. You upload your data, pay monthly, and hope the company is still around next year. These tools take the opposite approach. Each one looks for an AI model you already control and uses that: the Claude command line tool if you have a Claude subscription, the Codex command line tool for a ChatGPT subscription, an Anthropic or OpenAI API key, or any OpenAI compatible server, including local models through Ollama or LM Studio.
That design has a practical consequence: the AI features cost you nothing beyond what you already pay, and everything that can work without a model still does. labparse parses every known format offline, statsage produces its full statistical report with no model configured, and Literature Radar only needs the model for ranking. The AI makes each tool better, but never holds your work hostage.
All three are MIT licensed, and the source for each lives on our GitHub. Read it, fork it, fix it, or file an issue if something in your lab does not work the way it should.
Literature Radar: Never Miss a Relevant Paper
Keeping up with the literature is a job nobody has time for. Literature Radar is a small self-hosted app that does the sweep for you. You describe your research interests in plain language, add a few topic queries, and a scheduled job polls arXiv, PubMed, Crossref, bioRxiv, and medRxiv. Your AI model scores each new paper from 0 to 10 against your interest profile, with a one line reason, and on your chosen schedule you get a single email digest with the best papers first.
It is one PHP app with one SQLite file, so it runs on nearly any server, home machine, or Docker host. The Literature Radar guide covers setup, sources, and the admin area in detail.
labparse: Tidy Tables from Any Instrument Export
Every plate reader and spectrometer exports its own hostile file format, and most labs quietly waste hours writing throwaway parsing scripts. labparse is a Python package with one job: point it at the file your instrument produced and get back a tidy table, one row per reading, with well, row, column, cycle, time, and value. Known formats parse instantly from bundled recipes. Unknown formats are shown to your AI model once, the model writes a recipe, the recipe is verified against your actual file, and from then on that format parses offline with no model involved.
It installs with pip and works from Python or the command line. The labparse guide covers the recipe system, bundled formats, and the output schema.
statsage: The Right Statistical Test, Every Time
Picking the wrong statistical test is one of the most common mistakes in published research. statsage is a Python package that takes your table and the columns to compare, checks normality, variances, sample sizes, and outliers, then picks the appropriate test and runs it properly, with effect sizes, confidence intervals, and post-hoc comparisons where relevant. It draws a publication quality figure and writes the methods paragraph and results sentence you can paste into a manuscript.
Every report shows the reasoning path that led to the chosen test, so you learn the statistics while getting the answer. It works completely offline, and if you have an AI model available it polishes the written narrative without ever being allowed to change a number. The statsage guide covers everything it supports.
Getting Started
The two Python tools install in one line each:
pip install labparse
pip install statsageLiterature Radar is a clone and configure setup, with a Docker option for one command deployment. Each tool's guide walks through it step by step, and each GitHub repository has the same documentation alongside the code. This collection will keep growing, so check back, or watch the GitHub organization to see new tools as they land.