Turn an experiment plan into steps and parameters you can follow: save it from the AI you already use, or paste text or upload a PDF. Every parameter shows its source, and you check them before you start.
Each one maps to something you can use today.
Each imported field is labeled stated, partly stated, inferred or missing, and nothing becomes an experiment until you confirm it.
How review worksDilutions, molarity, cell seeding and centrifuge speed use fixed formulas, and you are told to pre-dilute when a volume falls below 1 μL.
Open the calculatorsExperiments are split into sub-experiments, steps and parameters. Change an approach and save it as a new version, with v1 and v2 kept side by side.
How experiments are organizedEach run records observations, deviations and a conclusion. Change a parameter and the sub-experiments that use it are flagged for review.
How run logs workVara does not make scientific calls for you. It keeps the plan you worked out with your AI so you can follow it at the bench and record every run.
Discuss the plan in Claude, ChatGPT, Doubao, or whichever AI you use. Once it is connected, it can also read the experiments already in Vara.
Your AI saves the experiment to Vara as a draft with a source on each parameter, and it becomes an experiment once you confirm it. It can also put the work on your experiment schedule.
Follow the steps at the bench and record observations and deviations. Your experiment schedule can be subscribed to in iPhone, Google, or Outlook Calendar.
AI apps that can connect to Vara
All of them connect over remote MCP. Vara is not listed in the ChatGPT app directory yet, so for now you add it by hand from Plugins in ChatGPT.
See the setup guideNo connector? You can also paste text or upload a PDF, Word file, or photo, orcopy a prompt into any AI and paste the reply back.
The experiment group shows the whole workflow, and each sub-experiment holds its procedure and parameters. Experiments your AI saves sit here next to the ones you build yourself.
Human organoids with an autologous tissue-resident immune compartment
Recaldin et al · Nature · 2024
Sampling & ethics
T_RM isolation
Gut organoids
IIO co-culture
mIF staining
Live imaging
scRNA-seq
Drug induction
Try asking Vara:
Graduate Students
Running a paper's protocol for the first time and need to know which values the paper states and which you have to decide
Researchers
Managing multiple experiment versions and tracking changes across iterations
PIs & Lab Leads
Give students a structured workflow to reduce back-and-forth
Over remote MCP, apps like Claude and ChatGPT can read and write your experiments, steps, run logs, and schedule. You can revoke access in Settings at any time.
Experiment groups, sub-experiments and procedure steps in three levels, so you can see where each experiment stands.
Define variables and formulas and link them to steps. Change one parameter and the values tied to it update.
Record observations, deviations, and a conclusion for every run, so the full history of an experiment can be traced.
Generate a link in Settings and your experiment schedule shows up in iPhone, Google, or Outlook Calendar. It is read-only and follows your plan.
Paste text, upload a PDF, Word file, or photo, copy a prompt into any AI and paste the reply back, or import a public protocol from protocols.io.
Dilution, molarity, cell seeding, RCF to RPM and plate layouts. Free, no login.
What to record and how, with a printable lab notebook template.
Western blot, qPCR, cell culture and more: what papers leave out and what to check first.
Who Vara, Benchling, eLabFTW, protocols.io and others each fit.
We built privacy into the architecture, not on top of it. Four guarantees below.
Platform administrators can only see experiment group names and aggregate statistics. They cannot access your experiment steps, parameter values, run logs, or any specific content.
Each user's data is fully isolated at the database level. Every query checks ownership, so no one else can access your data.
Vara never uses your experiment data to train models. The built-in AI sends only what a request needs to the model provider, and only when you use it. When you connect Vara to Claude, ChatGPT or another assistant, that conversation follows the provider's own terms.
Traffic is encrypted over HTTPS, passwords are stored as bcrypt hashes, files sit in private storage behind server-side access checks, and sessions use HttpOnly secure cookies.
We're in beta. We read every email:
support@varaos.com