BeakerBot
Ask BeakerBot to work with your data, sequences, papers, and experiments inside ResearchOS. Use the app’s tools directly from the conversation.
AI is optional. The local app is free. See pricing for current beta access and separate AI usage charges.
Inside the app
Ask for a project, request a different name, approve the revised proposal, and open the saved result.
What BeakerBot can do
Join tables, filter rows, and reshape data. Run statistical tests, fit curves, and make editable plots with ResearchOS analysis tools. You can inspect the generated code.
“Join these tables by sample ID, then compare the treatment groups.”
Fetch sequences from NCBI, extract annotated regions, design primers, and plan Gibson or restriction-ligation assemblies. ResearchOS sequence tools calculate the resulting constructs.
“Extract this feature and assemble it with my vector using Gibson.”
Generate tree-building commands and an environment file to run on your computer. Work with saved trees and find tables that match their tip names.
“Make an IQ-TREE workflow for my sequences.”
Attach a text PDF and draft a summary or extract its method into your library. Review the draft before saving it. Search Europe PMC for related papers.
“Extract the method from this paper into a draft I can review.”
Search records and full note text. Check overdue tasks, low stock, or purchases across a date range. Counts and totals come from your records.
“Show overdue experiments and low-stock inventory for this week.”
Create projects with linked experiments, attach methods, and schedule preparation tasks. Review the proposed work before saving it.
“Set up an experiment from this method with its preparation tasks.”
Preview changes to colors, labels, fonts, and figure dimensions. Apply the version you prefer, or undo the change.
Calculate power or sample size for two-group, paired, ANOVA, or correlation studies. Enter your assumptions and check the result.
Find compounds in PubChem, import structures into your library, and run your saved custom calculators with the inputs you provide.
Save supported steps as a macro, change their inputs, and run them again from the slash menu. Changes still require your approval.
Draft notes, update methods, organize projects, and log purchase or inventory changes. Review the changes before saving.
Ask where a feature is. BeakerBot can read the current page, navigate to another page, and highlight a control.
Open source, with clear limits
Write notes, analyze data, and make figures without using BeakerBot. AI has separate usage pricing and beta availability.
AI pricingResearchOS is open source under AGPLv3. BeakerBot defaults to GPT-OSS-120B, an open-weight model released under Apache 2.0.
Read the sourceYour notebook is stored locally. Hosted AI receives your message and the context used for the request. Open weights alone do not make processing private.
Model and privacy detailsThe current source defaults to GPT-OSS-120B hosted through Fireworks. Chat, image, and formatting requests can use different configured models. Open models give us more choice in how we run AI and help us keep costs low.
GPT-OSS model cardHosted requests send messages and relevant notebook context to the configured provider. This is different from running a model on your computer. Fireworks says it does not train on API inputs without explicit opt-in.
Provider privacy policyBeakerBot prepares the commands, installation steps, and environment file. You run the tree-building software on your computer.
Text PDFs can be attached for reading and draft extraction. Image attachments and matching a paper figure’s appearance require image support to be enabled and a suitable model to be configured. Matching a tree figure changes the styling of your own tree, not its topology.
You can review actions step by step or approve a plan. Requested analyses and plots can run directly. Note drafts and record changes use review controls, and destructive actions require confirmation. Available tools also depend on the workspace and your access.
It calls ResearchOS analysis engines and explains their results. You can inspect the saved analysis in your table. Review its explanation as you would any AI response.
Made by academics, for academics. Free for local use.