Researcher
Pulls primary sources, attaches a citation chip to every claim, surfaces prior work and the novel angle.
An instrument for experiment designFrom hypothesis to fundable plan
From hypothesis to fundable plan
in three minutes.
Frame a scientific question. Four agents draft a citation-grounded protocol — every claim cited, every reagent linked, every test green before the document is called ready.
01The problem
Contract research is slow, expensive, and ungrounded — most experiment plans cite nothing, link to nothing, and arrive weeks late.
02The method
Four agents, working in parallel against the same plan document.
Sextant is not a chatbot. It writes — and watches itself write. Each agent owns a slice of the plan, every edit is grounded in a fetched source, and a validation grid blocks the document from being called ready until every test passes.
Researcher01
Surveys the literature, anchors every claim to a citation, flags prior art.
Skeptic02
Challenges assumptions, demands controls, marks novelty vs. duplication.
Operator03
Sources reagents, prices the plan, lays a feasible timeline with dependencies.
Compliance04
Enforces ethics, biosafety, and lab-specific rules. Blocks unsafe drafts.
Skeptic
Argues against the protocol. Forces sham controls, statistical power checks, and falsifiable success criteria.
Operator
Fills the materials table with catalog numbers, real prices, lead times, and a Gantt that respects dependencies.
Compliance
Runs ethics, biosafety, and your lab’s typed rules. The plan can’t be marked ready until this gate is green.
03The closed loop
Every correction becomes a typed lab rule. Applied to every plan after it.
When you correct a line in the protocol, Sextant captures the rule, names it, and re-applies it the next time a plan touches the same shape. The agents learn your lab — not the average lab.
Step 01 · Correction
You edit the protocol.
Inject 200 μL vehicle into n=8 mice; observe at 24h.
Inject 200 μL vehicle into n=8 sham + n=8 dosed; observe at 24h.
Inject 200 μL vehicle into n=8 sham + n=8 dosed; observe at 24h.
Step 02 · Typed lab rule
Sextant captures the pattern.
Rule · LR-12Always include a sham control arm in in vivo mouse studies, matched n.
applies to: study.species == “mouse” ∧ design.has_dose
Step 03 · Applied to next plan
Next mouse plan, drafted with it.
Vehicle injection arm: n=8 sham + n=8 dosed, observation at 24h.
— rule LR-12 applied automatically
— rule LR-12 applied automatically
Built on
AnthropicGeminiTavilyVercel
A plan that cites itself, prices itself, and learns your lab.
Three minutes from question to fundable. Every reagent links to a real catalog page. Every test must pass.
Open SextantValidation grid0 / 6 pass
Every reagent has a catalog URL
Budget sums correctly
No orphan protocol step
Citations resolve to real sources
Timeline dependencies valid
Compliance pipeline passes