Your AI drug-design scientist
Give it a research goal. It plans, calls real scientific tools, verifies its own findings, and returns a grounded, fully-cited result.
Not a chatbot. An agent that does the work.
Any target, disease, modality or species — it decides which tools to use, runs them for real, and builds on prior results.
Autonomous tool orchestration
It picks the right tool for each step and chains them — structure, variants, literature, databases, druggability.
Multi-agent parallelization
For large goals it spawns parallel sub-agents (e.g. profiling several targets at once) and synthesizes their findings.
Self-verification
Every answer is re-checked against the tool evidence. Fabricated accessions, scores or citations are flagged.
Long-running & memory
Runs persist and stream live, survive disconnects, and accumulate reusable knowledge across sessions.
It calls real tools — and shows its work.
No black box. Watch the tool-call trace in real time, then read an answer where every number traces back to a live tool call.
A growing toolbox of live scientific APIs.
Called directly — open and commercially-friendly engines wherever possible.
Grounded, cited, honest about uncertainty.
Every claim is traceable
Accessions, pLDDT, PMIDs and trial IDs come from real tool calls shown in the trace — not invented.
Flags what it can't support
A verification pass scores evidence-grounding and lists any claim the tools didn't back.
Put an autonomous scientist on the problem.
Download Catalyst AI and start your first run.
Download for Windows