Gosset Build on Gosset
Build your own agents with the infrastructure we use ourselves
You can access Gosset from our user-friendly web app,
or put Gosset to work directly inside your own stack.
API
Query our knowledge graph programmatically for drugs, trials, results, targets, companies, and deals. Get typed endpoints that drop into your own models and pipelines.
MCP
Our Model Context Protocol server plugs Gosset into Claude, OpenAI Agents, or any other MCP-compatible client. Your agents get real-time, structured pharma intelligence.
Tools & Models
Give your agents access to validated, accurate forecasts of clinical trial outcomes and asset valuations.
How we measure accuracy →Subagents
Drop pre-built Gosset agents into your system as subagents. Each handles a focused job like competitive landscaping or asset diligence.
The data
Backed by the most comprehensive pharmaceutical intelligence knowledge graph
Our always-on knowledge graph ingests trial registry updates, press releases, regulatory filings, company websites, conference abstracts, and publications, then structures their information.
Results no one else has
We extract safety and efficacy results from every trial that has disclosed them, whether in a paper, poster, press release, or investor deck. No other platform matches our coverage.
Structured, not scraped
The information in every source is structured and organized with detailed ontologies so that it is linked and instantly findable.
Always current
The graph updates in real time as new evidence is published, so competitive landscapes and risk assessments reflect what happened this week, not last quarter.
How we measure forecasting accuracy
Gosset develops models for clinical trial outcome forecasting and asset valuation. Our models are evaluated using out-of-sample, point-in-time data on trials and assets whose actual outcomes are now known. On trial completion date forecasting, our model predictions land a median of 47 days closer to the true completion date than the sponsor's own posted estimates. Probability-of-success estimates are reported with calibration curves so you can see how predicted probabilities track observed outcomes.
We'll walk through the methodology, the evaluation set, and the limitations in detail on a call. We'd rather show our work than ask you to take the number on faith.