| Built for | A search engine and agent that answers across genes, sequences, structures, drugs and metagenomes with one question | The human gene database: one card per gene, with many sources brought together | A platform that gathers evidence linking targets and diseases and scores it |
|---|
| How you ask | Natural-language conversation (it remembers follow-ups), a pasted sequence, identifiers and coordinates | Search by gene name or keyword, then read the card | Search a target or a disease, then read the scores and the evidence tables |
|---|
| Data scope | Human genes plus billions of metagenomic proteins, tens of thousands of non-human genomes, domains, structures, immunology and raw binding measurements | Centred on human genes, with more than 190 sources brought together | Target–disease evidence: genetics, somatic mutations, drugs, pathways, expression, animal models, literature |
|---|
| Crossing layers | Asks dozens of layers at once and points to where independent evidence agrees, marking a lead as a lead, not a discovery | Lists a gene’s many sources on one card | Adds up scores per kind of evidence into a target–disease score |
|---|
| Where you start | From a sequence, on to homologs, domains and structure; your own sequence is folded to 3D on the spot | From a gene name or identifier | From a target or a disease |
|---|
| Automation | One REST API and one MCP server reach dozens of layers, for Claude, ChatGPT, Cursor or your own agent | Not found in its public documentation | A GraphQL API and its own MCP server (Open Targets data only) |
|---|
| Terms | Free to try every month; company plans from $2,000 a month | Free for non-profit academic institutions and hospitals; companies need a licence | Public service with a public API |