Skip to main content
Doc IRBIS-2026Class Governed-AIEngines R·Python·StanLang EN·RU·KK

Your researchers already use AI.
Make it safe, provable, and yours.

Every AI chat with a spreadsheet attached is an unlogged data exposure. IRBIS gives your institution one governed workspace where AI assists real research — and never sees a raw row.

raw_rows_to_ai0
ops_audited100%
audit_logs1 / project
methods_reviewed90+
§02 · The exposure you can't see

When a researcher pastes patient data into a chatbot, your institution carries the risk.

An institutional IRBIS subscription replaces that habit with a governed alternative.

Audit readout
raw rows reach any AI0
of AI operations audited0%
audit log for every project0
§03 · Workflow

From field data to published finding — one governed workflow.

ST-01Collect · Field

Collect and organize, in one place

Surveys with consent and supervisor review. Datasets, files, and versions in governed project folders. Provenance attaches at the moment of collection.

ST-02Research · Literature

Study the literature before the analysis

The Research Hub brainstorms hypotheses with you and summarizes the available research — so every analysis starts from the evidence, not a guess.

ST-03Analyze · Workbench

Analyze with any AI agent — behind the firewall

Choose the agent that suits the task. Whichever model your researchers prefer, it works through the same vetted operations and never sees raw data.

ST-04Publish · Manuscript

The manuscript, rebuildable from scratch

Reports are built from deterministic analysis artifacts. Rerun the analysis and IRBIS rebuilds the manuscript — reporting whether results reproduced, and why.

§04 · The methods index

Depth your statisticians will respect.

Every method runs on authoritative engines — R, Python, Stan — never an LLM doing arithmetic.

DomainCoverageEngineStatus
Inference & regressiont-tests to mixed models, with assumption checks built inR / TS
Shipped
Epidemiology26 tools: survival, case-control, and cohort designsR
Shipped
Simulation & power38 scenarios and a study design wizardTS
Shipped
Econometrics & social scienceTool libraries for economic and social researchR / TS
Shipped
§05 · The firewall, in depth

AI operates the platform. Only the platform touches the data.

Closed operation surfaceThe AI calls only vetted operations — metadata, summaries, pivots, charts, and a catalog of reviewed tests. No free-form code reaches your data.
Classified columnsEvery column is labeled from public to sensitive. Confidential fields are aliased before anything reaches the AI.
Aggregates, never rowsResults pass through small-cell suppression and rare-category redaction. The AI sees counts and coefficients, not individuals.
Audited by defaultEvery operation is content-hashed and logged with its policy result and dataset version.
Reasoning agent
Fast drafting agent
Budget agent
Different agents, one firewall — every model works under the same rules.
§06 · Proof for the office of research

No installs, no raw rows out, no language barrier.

Browser-firstNo installs, no IT burden
EN / RU / KKTrilingual reporting
Data-law alignedBuilt for local compliance
Vetted methods90+ reviewed statistical tools
§07 · Plans

Plans for every institution

Tier 0

Free

calculators

Statistical calculators, open to everyone

Tier 1

Starter

no-AI

The full platform without AI — files, engines, reports

Tier 2

Basic

light-AI

Light AI assistance on budget models

Tier 3 · Most popular

Pro

manuscripts

Regular analysis with manuscript help

Tier 4

Premium

thinking-tier

Heavy use with thinking-tier models

Starter through Premium are institutional seat plans. Per-seat pricing is shared on request.

Ready when you are

Bring governed AI research to your institution.

Request a demo