Litmus Science
Lab Execution for AI-Driven Science
Litmus turns scientific questions into executable research programs. Describe a hypothesis and the copilot drafts a rigorous experimental plan, pressure-tests feasibility and statistical rigor before anything runs, then routes execution through a vetted network of outsourced labs. Results return structured, and the copilot proposes the next experiment — closing the loop from data back to hypothesis.
Who Litmus Is For
- Virtual biotechs — Validate platform hypotheses and run many experiments in parallel without building a full biology team
- Academic researchers — Grant-friendly workflows built around funding and publication timelines, with Delayed Release or Open Access IP tiers
- Program managers — Centralize quotes, protocols, and status across every lab so one person can manage what used to take a team
- AI agents — A lab in the loop. Agents can design and submit real experiments through Litmus using the open-source Litmus MCP server
How It Works
- Design — Describe the scientific question. The copilot drafts a rigorous plan: models, endpoints, controls, and powering.
- Validate — Feasibility, cost, and statistical rigor are pressure-tested before anything runs, catching underpowered designs early.
- Execute — The study runs through the right labs in the vetted network, with standardized workflows and full documentation.
- Interpret — Results return structured. The copilot analyzes them and proposes the next experiment.
Key Features
Experimental Design Copilot
Describe a hypothesis in plain language and get back a structured plan — models, endpoints, controls, and statistical powering — pressure-tested for feasibility before anything is scheduled.
Vetted Lab Network
One point of contact across preclinical assay types — in vivo, structural biology, protein science, and molecular/bioanalytical work — with quotes compared across labs including Gingko Cloud Lab, Sword Bio, and iHisto.
Structured Results
Every assay returns structured JSON/CSV output built for computational modeling, not a PDF someone has to retype. Results reference the hypothesis that produced them, so the copilot can propose the next experiment automatically.
MCP Server for AI Agents
An open-source MCP server (opens in a new tab) lets AI agents design and submit real experiments programmatically, the same way they'd call any other tool.
Assay Menu
All assays return structured JSON/CSV output built for computational modeling.
| Assay | Turnaround | Output |
|---|---|---|
| ELISA (Protein Quantification) | 3–5 days | 4PL curve fitting, raw + processed JSON/CSV |
| qPCR (Gene Expression) | 2–5 days | 384-well, ΔΔCt method, full Cq raw data |
| CTG (Cell Viability) | 3–5 days | Automated dose-response fitting, IC50 with 95% CI |
| SPR (Binding Kinetics) | 5–7 days | Label-free kon/koff/KD, sensorgram + kinetic fit CSV |
| Cell-Based (Phenotypic & Imaging) | Custom | High-content imaging, multiplexed endpoint readouts |
See the full assay menu (opens in a new tab) for specs and use cases.
Pricing
Monthly subscription, no annual contract, cancel anytime.
| Plan | Price | For |
|---|---|---|
| Free | $0/mo | Individuals exploring the platform — 1 active experiment at a time |
| Developer / Starter | $299/mo | Unlimited active experiments, unified umbrella MSA & consolidated billing, agent guardrails & spend controls |
See pricing (opens in a new tab) for full details.
Quick Links
Data Privacy & Compliance
- Litmus does not store, sell, or share research data. Protocols, results, and IP stay private and belong to the customer by default, with no exceptions.
- BSL-1/BSL-2 only — No BSL-3/4 work
- No controlled substances — DEA-scheduled compounds prohibited
- No human subjects — In vitro only
- Lab vetting — Every lab in the network is vetted before it can accept work
- Conservative review — Uncertain requests declined; appeal available