An agentic multi-agent studio for ethical animal research at <Institutional Name Redacted>. Drafts, reviews, surfaces decisions. Humans approve.
A support tool for the people who do the work. Built to accelerate protocol development, IACUC compliance, vivarium operations, study design, data integrity, and reporting — without ever replacing veterinary, ethical, or scientific judgment.
Replacement, Reduction, Refinement gate every protocol, study, and procedure. Not a checkbox — an architectural commitment, with a Welfare Lead holding veto authority.
No autonomous action on live animals or live systems. The studio drafts and reviews; humans submit, sign, and execute. Always.
Built for <Institution Redacted>. Pediatric relevance is a first-class review criterion in every study design, not an afterthought added at publication.
Departments rename. Departments dissolve. Stable IDs and versioned name-history registries keep legacy data attributable across decades.
Russell & Burch, 1959. The ethical foundation of humane experimental technique. Mandatory in every protocol. Veto-enforced.
ASKED FIRST, EVERY TIME.
Substitute non-animal alternatives whenever scientifically valid. In vitro systems, iPSC-derived models, organoids, organ-on-chip, computational models, less sentient species.
FEWER ANIMALS. BETTER SCIENCE.
Obtain the same or more information from fewer animals. Power analysis with stated effect-size sources. Factorial design. Repeated measures. Shared controls. Pre-registration. Allocation-aware.
DECREASE INCIDENCE. DECREASE SEVERITY.
Refine procedures, housing, handling, endpoints, and lifetime welfare. Species-appropriate enrichment. Multimodal analgesia. Objective humane endpoints. Refined techniques. Pediatric sensitivity windows.
Each artifact passes through the right hands. Tier I sets policy. Tier II owns workflows. Tier III executes. The 3Rs & Welfare Lead can veto anything, anywhere in the chain.
Strategic and compliance oversight. Final reviewers. Policy authority.
Workflow ownership. Primary authors and reviewers within their domain.
Task execution. Narrow scope. Deep focus. Report up to a Lead.
Type slash. Pick a skill. The studio routes the request to the right agents and produces the artifact. Eight skills are fully implemented; twenty-two are scaffolded for first-use expansion.
/start/team-review/compliance-audit/project-status/study-design/protocol-draft/iacuc-draft/3rs-review/power-analysis/humane-endpoint/welfare-monitoring/randomization-plan/blinding-protocol/endpoint-criteria/vivarium-plan/husbandry-sop/enrichment-audit/census-manage/environmental-monitor/supply-inventory/surgery-checklist/anesthesia-protocol/data-dictionary/stats-plan/analysis-pipeline/report-generate/publication-draft/socratic-explain/sop-generate/training-moduleEvery in vivo study faces two independent ceilings. Operational capacity is the lesser of the two.
// effective concurrent capacity capacity(species, PI) = min( protocol_authorized_count, arcac_allocated_cages × cage_density(species) )
Two operational. One under construction. ~30,000 cages active program-wide.
Mouse-dominant census. Population proportional to a multi-PI, multi-program, AAALAC-accredited research operation.
Stable internal IDs. Name-history registry. Renames don't break legacy data attribution.
NO REAL DATA.
NO REAL IDS.
EVER.
No animal IDs, ear-tag numbers, MRNs, PI names, or grant numbers in committed files. Placeholders only. Hooks warn at write-time.
THE STUDIO DRAFTS.
HUMANS APPROVE.
No autonomous action on live animals or live systems. The studio produces drafts and reviews; institutional submission is always a manual human action.
THE 3Rs & WELFARE LEAD
CAN VETO.
A welfare-substance veto from the 3Rs & Welfare Lead is binding. No other agent's approval can override it. Concerns are addressed; not bypassed.
$ git clone https://github.com/Vandevious/animal-research-ops-studio
$ cd animal-research-ops-studio && claude
> /start