ki27_cell_study · Outputs
OrcStat Outputs view with APA 7 tables for a two-way ANOVA
New: Pro Team for labs and departments See plans

Statistical analysis,
orchestrated by AI.

Ask a research question and get a validated analysis, from assumption checks to an APA-ready report. The AI runs on your own hardware.

ki27_cell_study.sav

14 days free, no card needed

Opens the files you already have, with labels and missing codes intact.

SPSS .sav.zsav.porStata .dtaSASExcelCSVParquetJSONSQLiteDuckDBOpenDocumentSPSS .sav.zsav.porStata .dtaSASExcelCSVParquetJSONSQLiteDuckDBOpenDocument

How it works

From raw data to a defensible result in three steps.

Bring in data. Open .sav, Stata, SAS, Excel or Parquet with labels, missing codes and measurement levels intact.

Ask your question. The orchestrator plans the analysis, checks assumptions and picks the right test when one fails.

Report and reproduce. APA 7 tables and a results section in Word, with a syntax journal that re-runs everything.

The model writes the words. The engine computes the numbers. Every statistic comes from a validated engine, and every number the AI writes is checked against the output before you read it.

The OrcStat ruleBuilt into every AI feature

The orchestrator

A statistician on call, on your own machine.

  • Plans that fit your design. Analyses chosen from your question, variable types, sample size and design.

  • Checks built in. Normality, equal variances and outliers are tested for you, with the right alternative when one fails.

  • Explanations you can verify. Every number in the AI's text is checked against the engine's output before you see it.

Try the orchestrator
You

Does kinase activity differ by treatment, and is it the same in both cell lines?

Analysis planLocal · 8B
  1. Profile9 variables, metadata only
  2. CleanMark 999 as missingNeeds approval
  3. CheckNormality, equal variances
  4. TestWelch ANOVA, Games-Howell
  5. ModelTreatment × cell line
  6. ReportAPA 7 results to Word
Kinase activity differs strongly across treatments, F(3, 140) = 125.81, p < .001, η² = .73.3 numbers verified against output
Remove stepApprove and run

The workbench

Everything a careful analysis needs, in one place.

Bring your own model

Any local model. OrcStat talks to the model servers you run and scores each one on statistics tasks.

Ollamallama.cppLM StudiovLLMOpenAI-compatibleYour GPU serverLaptop CPUAir-gapped
Report writer

APA 7, linked to the output. Methods and results in Word, each number tied to its table.

Results

A one-way ANOVA showed kinase activity differed across treatments, F(3, 140) = 125.81, p < .001, η² = .73.

Data Editor

Ten million rows on a laptop. Data View and Variable View, the way you're used to.

OrcStat Data Editor with a labelled dataset
Syntax journal

Every click, replayable. Dialogs and AI actions become syntax.

Dialog: DescriptivesRecorded 14:05
AI: ONEWAY, approvedRecorded 14:09
Re-run all
Pro Team

One plan for the whole lab. Owner, admin and billing roles, invitations by email or link.

ALMKJRSP +

Every module is included. No add‑ons to buy.

Chart Builder Engine-computed charts you can edit and export.
Encrypted projects AES-256 projects and password-protected .sav files.
Python extensions Programs, custom dialogs and batch jobs.
Complex Samples Design-correct estimates from survey data.
Multiple imputation Pooled results across imputed datasets.
Bayesian statistics Bayes factors and credible intervals alongside p values.
Custom tables Publication tables with nesting and stacking.
Browser version The full engine in your browser through WebAssembly.

Privacy and security

Your data stays put.
The analysis comes to it.

The engine runs on your computer, or in your browser. The AI talks only to model servers you configure. Nothing about your data leaves the building.

Local Only
AES-256
Works offline
No telemetry
Audit log

You choose what the AI sees. By default, variable names, labels and output tables only. Raise or lower it per project and preview every request.

Models on your hardware. Ollama, llama.cpp, LM Studio or a server on your network. Local Only mode blocks every other destination.

A licence check, nothing more. Four things: licence ID, a hashed device ID, app version and operating system. Never your data or analyses.

How OrcStat handles your data, for security and IT reviews

Under the hood

Depth you can measure.

148procedures, every module included
6,700+automated tests on every change
10Mrows, comfortably, on a laptop
0bytes of your data sent to a cloud

Plans

Every module, in every plan.

You pay for the private AI and the controls, never for statistics add‑ons.

Pro Team

Pro for labs, departments and firms, with shared billing and roles.

$40/ user / month Billed annually. $35 per user at 50+ seats
Create a team

Buy seats, then invite people by email or link

  • Everything in Pro Personal
  • Owner, admin and billing roles
  • Invite by email, CSV or join link
  • One invoice for the whole team
  • Lower price per seat from 50 seats
Institution

Pro for every seat, with the controls IT and governance teams need.

Custom Site licences and seat packs, usually annual
Contact sales

Pilots, security reviews and procurement

  • Single sign-on and user provisioning
  • Invoice or purchase order
  • Admin policy file and Local Only enforcement
  • Offline activation for air-gapped machines
  • Course groups for teaching
  • Priority support and onboarding

Compare every feature, student discounts and the beta offer

Validation

Linear regression and univariate statistics reproduce the certified values of the NIST reference datasets, on every change.
Golden test suiteRun in CI on every commit
NISTStatistical Reference Datasets
6,700+automated tests
100%of statistics computed by the engine, never the AI

Use cases

Built for people who answer to reviewers.

Students, researchers and clinical teams use OrcStat to turn raw data into results they can defend.

Explore use cases

Questions, answered.

Privacy, models and files. More on the pricing page.

Does the AI see my data?

By default it sees variable names, labels and output tables only. You can raise or lower the exposure level per project, and preview every request before it is sent.

Which models does OrcStat work with?

Anything served by Ollama, llama.cpp, LM Studio, vLLM or another OpenAI-compatible server on your machine or network. OrcStat scores each model on statistics tasks.

Can the AI get a number wrong?

The AI never calculates. A validated engine computes every statistic, and each number in the AI's text is checked against the output before you see it.

Will my SPSS files and syntax work?

OrcStat reads and writes .sav and .zsav, reads .por, and runs a large subset of SPSS-style syntax. The copilot converts the rest.

Do I need a GPU?

No. The engine runs on any laptop. For AI, an 8B model runs well on 16 GB of RAM, and OrcStat recommends a size for your hardware.

What happens after the trial?

Subscribe to keep going. If you don't, OrcStat switches to read-only mode: you can still open, view and export your projects, but not run new analyses or use the AI.

Every analysis, ready to re‑run.

Save the project, share the syntax, or export a self-contained report. Anyone can reproduce your numbers.

14 days free, no card needed