Design an experiment, run it, and walk away with a record anyone can recompute.
Illustrative example: a scripted preview (7 of the 12 sections shown), not a live run.
Design an experiment, run it, analyze the results, and leave behind a permanent, verifiable record of exactly what happened. Click through the six steps below, or let it play.
Illustrative example: a scripted preview of the product, not a live run.
Tell Olto what you want to find out in plain language. It drafts a full, 12-section protocol in about a minute, then you edit it as a drag-and-drop step graph.
Editable step graph · versionedNot an AI opinion. It is a deterministic rigor score across seven reproducibility signals: controls, replication, sample size, statistics, quantitative parameters, randomization, and safety. Run the same protocol through it twice and the number is identical. It gets a public URL and a citation string, and there's no login to read it. There are 104 live in the public library right now.
Researchers describe the same experiment in completely different words. Olto converts each protocol into its underlying experimental design (units normalized to SI, methods and controls extracted) and content-addresses it into one stable, recomputable identity. Different wording, same design, same fingerprint. Deterministic, no AI.
Add 5 mL DMEM. Incubate at 37 °C for 30 min. Untreated controls, n=3. Analyze by t-test.
Pipette 0.005 L DMEM. Incubated at 310.15 K for 1800 s. Vehicle control, n = 3. Student's t-test.
rpf_p78hs7mthhhmz409jx0kacfby8Identical experimentLike Olto's statistics workbench and notebook Python, the fingerprint is computed entirely in your browser, so your rawest data and protocol text never have to leave your machine.
Every public protocol carries four design-quality scores from its AI review: feasibility, controls, reproducibility, and clarity. This is how those scores distribute across the public library today, computed live rather than illustrated. The library is seeded by the Olto team while it opens to outside authors. (The deterministic seven-signal rigor score, which no model touches, lives on each protocol's passport.)
Drafting a first protocol is a language problem, and a model is good at it. Deciding whether a design is rigorous, whether two methods are the same method, or what a variant classification is are not language problems, and those run as ordinary functions you can re-run yourself.
The other 423 compute. You do not have to take that on trust: the engine registry lists every sealed engine with its citation, its version and a content hash, and re-runs all of them each time the page loads. The rigor score is specified in full, so you can work it out by hand.
One exception, stated rather than buried: the protocol risk assessment is produced by a model, not computed, and is labelled as an assessment wherever it appears. Prompts and uploaded files are sent to our AI provider, so Olto is not end to end encrypted. The full system card lists every subsystem and whether its output is retrieved, computed, extracted or inferred.
Every claim below is a shipped behaviour with a specific mechanism behind it, and every one of them is something you can check on a real protocol in the public library before you sign up.
The same reproducibility-first core, framed for three audiences. The workspace adapts its navigation, terminology, and tools to match the track you pick.
The full life-science lifecycle: AI protocol design, guided test runs, a client-side Statistics Workbench, literature synthesis, and a citable Reproducibility Passport.
Try the research workspace →A deterministic workbench: 32 calculators, tolerance stack-up, multi-level BOM, GUM uncertainty and metrology. Every number shows its formula, runs in your browser, and is never metered.
See the engineering workspace →An inquiry-science workspace with an AI protocol coach under teacher oversight, a weighted gradebook, quizzes, discussions, and verifiable certificates, all in a school-controlled, privacy-conscious space.
Explore Olto for Education →One connected platform for the whole of experimental research: design, run, analyze, document, and collaborate, for life sciences and engineering alike. Everything below is included on paid plans. AI research actions are metered; storage scales by plan; features are not.
Flip from the life-sciences lab to the engineering lab and the navigation, terminology, and tools adapt. The engineering side is deterministic all the way down: every number shows its formula, runs in your browser, and is unit-tested against the published standard. No AI anywhere near the result, and it's never metered.
The block notebook works in both modes: drop a deterministic calc inline beside your Python, sign it, export. Read the engineering docs →
Describe a goal, get a 12-section, publication-structured first-draft protocol in under a minute, scored on feasibility, controls, reproducibility, and clarity. Edit inline, export to PDF, fork to anyone.
A tool-using assistant that searches your own protocols, papers, and inventory, and can run a statistical test, answering grounded, with citations and an audit trail.
A block-based electronic lab notebook that runs Python in your browser. Run a t-test, get a matplotlib figure inline, then sign, version, and export the entry to PDF.
Run real t-tests, ANOVA, correlation, regression, and non-parametric tests entirely client-side, so your raw data stays in your browser unless you choose to save or export it. Effect sizes, CIs, and plots included.
Execute any protocol as a step-by-step session. Per-step timers, pass/fail/deviation logging, findings and evidence captured as you go, then a session-report PDF.
A reference manager with collections, full-text search, Ask-AI Q&A grounded on a paper, multi-paper synthesis, and 13-style citation export.
Order lab tests, track results, auto-flag against reference ranges, release a Certificate of Analysis, and invoice. On Lab plans and up.
Deterministic computer vision in the browser: lane densitometry and cell counting. The numbers are yours; AI is used only for QC.
Map an experiment to the ISO, ASTM, USP, and ICH standards it should meet, and surface them right on the protocol.
Presence, live cursors, and inline comments. Co-edit protocols and projects with version history and approvals.
Track stocks, expirations, locations, suppliers, and lot numbers with a full transaction ledger. Get warned before you run out, not after.
Browse a rigor-scored reference library we wrote and scored ourselves. Fork any one to start from something real, and put your name on yours.
Start free, no card. AI is metered; on paid plans everything else is unlimited (free Explorer includes 5 protocols).
We built Olto because we wanted to use it. The alternatives were either too expensive or too narrow. This comparison is made in good faith. If something looks wrong, tell us and we'll fix it.
These are specific, checkable capabilities that set Olto apart: each one is in the product today, not a roadmap promise.
For context, enterprise lab platforms are typically priced through sales rather than published; Olto publishes its pricing openly at $79–$499 / month. The tools below serve overlapping but different jobs, so here is a neutral one-line description of each.
Competitor capabilities and pricing were reviewed on August 2026 and may change; verify current information directly with each provider. This comparison is made in good faith. If something looks wrong, tell us at support@oltodiscovery.com and we'll fix it.
Your statistics and notebook Python run in your browser, so sensitive data can stay on your machine; everything else is encrypted in transit and at rest and isolated per-tenant by Postgres row-level security. See exactly what we do (and don't) claim →
Generate a real, rigor-scored protocol right now. No account, no card, no demo call. Then keep it free.