FAQ

Vara is an AI research assistant for wet-lab scientists, available on the web and macOS. On iPhone and Android you can add the web app to your home screen (native mobile apps are coming soon). It turns an experiment plan, protocol, or paper into an executable, structured workflow with linked parameters, run records, and AI-assisted failure review. It is built for graduate researchers, working scientists, and lab PIs. As of July 2026, 1,000+ beta researchers have created 900+ real experiments.

Break the culture workflow into parameterized steps (seeding density, passage timing, reagent concentration), run them on a timeline with timers, and log results and deviations per run. Changing one parameter flags dependent downstream steps for recalculation, and iterations across versions are tracked.

Drop the paper's PDF into Vara. It extracts the method into parameterized steps and labels each field with a source-confidence level (explicit / partial / inferred / missing) linked to the original text. After you verify, you run it, and can use AI to review failures.

Vara labels every AI-extracted field with a source-confidence level: explicit, partial, inferred, or missing. The model extracts and you verify, rather than trusting it blindly. Calculations between parameters run on a deterministic engine, not AI guessing.

No. Vara imports plans you already discussed in ChatGPT, DeepSeek, or Claude: generate them in Vara's prompt format and paste them back for one-click parsing. You can also upload a protocol as PDF, image, or SOP. This is designed to avoid cold-start data entry.

Vara supports discipline-specific reproduction across 10 disciplines, handling code, formulas, and data fields differently per field. Content currently focuses on wet-lab disciplines: life science, biology, and chemistry.

Vara has a free tier that manages 3 experiments. Paid plans are Pro at $6 per month and Pro+ at $19 per month, billed in USD and processed by Creem; see the pricing page. Bring-your-own-key (BYOK) for language models is also supported.

One wet-lab researcher reported that after managing and reproducing experiments in Vara, trial-and-error and repeated runs dropped by about 60%. Linked-parameter recalculation, source-confidence labels, and failure review are the main reasons for fewer wasted repeats.

Notion is generic notes. Benchling is enterprise ELN for biology. Vara is specifically for wet-lab researchers, AI is first-class (parse PDFs / edit experiments directly), data is structured (params + steps + materials), not document blobs. Fastest onboarding for solo researchers.

Usually transient upstream issue (Volcano Ark / MiniMax occasionally hiccup). Vara auto-retries once, after which it shows Upstream AI temporarily unstable, please retry. If persistent, switch model (ModelSelector bottom-right).

The current model has no vision. In the model selector, pick one with vision support (e.g. Doubao Seed / MiMo / Kimi), then re-upload.

Check Settings → Appearance → AI output language. Pick "Force Chinese" or "Follow system" (with Chinese UI). AI output language is independent of UI — you can use Chinese UI but English output for sharing with overseas collaborators.

All data lives in a managed cloud database (Asia-Pacific / Singapore region), with automatic periodic snapshot backups. The macOS client edits offline and auto-push on reconnect. Local backup: Settings → Export data (feature pending).

No. Quota is a sliding window — once your earliest charge slides out of the 5-hour / 7-day window, part of it restores automatically. No need to wait for month start.

Yes. Pro ($6/month) and Pro+ ($19/month) can be bought on the pricing page; payments are processed by Creem and you can cancel anytime. During the beta, completing the survey or using an invite code also gets you 30-90 days of Pro for free. Beta users who completed the survey on or before October 7, 2026 (Beijing time) get 10% off their first subscription, applied automatically at checkout.

Check Web first (Web is source of truth). If client is newer, push to Web. Stuck syncs: email [email protected] with account email.

Settings → [Feedback] button or email [email protected]. Include: steps to reproduce + screenshot + browser/OS + email.