Database & lookup tools
Instant, inline lookups against external databases — compounds, targets, literature, and trials — that return directly in the canvas with no job to track.
Database and lookup tools are the fast half of the catalog: instant, inline queries against established external databases. When the orchestrator calls one, it returns results directly in the canvas — there is no job to watch, no queued → running → completed progress card, and the credit cost is minimal. Reach for these to pull known data rather than compute something new: measured bioactivity, compound structures, protein annotations, published literature, purchasable suppliers, and clinical trials. This page is the front door to that catalog.
You reach a lookup by asking Ivy in chat, in plain language: "Find ChEMBL compounds with IC50 under 100 nM against EGFR", "Pull the UniProt entry for KRAS". Ivy picks the database, runs the lookup inside that turn and answers with the result. There is no job, no submit form and no CLI or MCP command for a lookup. UniProt and PubChem also have a read-only REST lookup.
No job to track
A lookup tool answers inline, usually in a few seconds. Nothing is dispatched to the GPU fleet and no jobs record is written — the result renders straight into the conversation as a canvas card. See How tools run for how this differs from a tracked Modal GPU job.
The catalog
ChEMBL
Search Europe's largest open bioactivity database for compounds, targets, and assay data.
PubChem
Look up PubChem by CID, name or structure and export CSV or SDF.
PubMed
Search PubMed/MEDLINE for biomedical literature using natural language or Boolean queries.
UniProt
Search the UniProt protein database for sequences, annotations, GO terms, and cross-references.
PDB
Search and retrieve protein structures from the RCSB Protein Data Bank.
ZINC
Query, filter, and download ZINC compounds for screening and docking.
Clinical Trials
Search ClinicalTrials.gov by condition, intervention, phase, and status.
Biotite's sequence fetch is not on this list: it pulls FASTA by UniProt accession, PDB ID or gene name, but it runs as a short tracked job rather than an inline lookup. See Biotite.
How these run
Database and lookup tools are the instant execution class. Because they query an external database over HTTP and return inline, they have the lowest latency, no artifacts to revisit, and minimal credit cost.
- How tools run — the execution-class explainer: what makes a tool an instant lookup versus a tracked job, and where each output lands.
Read next
Was this page helpful?