Small Molecule Discovery (ToolUniverse Claude Skill)

A ToolUniverse agent skill that takes a small molecule from identity resolution through bioactivity, drug-likeness, predicted targets, and commercial availability.


Type	Claude Skill (one of ToolUniverse’s pre-built agent skills)
Supplier	Zitnik Lab, Harvard Medical School
Availability	GA — part of the ToolUniverse skills collection (`skills/tooluniverse-small-molecule-discovery/`)
Pricing	Free / OSS (Apache-2.0); wraps public APIs (PubChem, ChEMBL, BindingDB, ADMET-AI, SwissADME, SwissTargetPrediction, eMolecules, Enamine)
Capabilities	Read-only — drives ToolUniverse tool calls; no data writes
Verified	works · 2026-07-20
Security	cleared · 2026-07-20 — provenance matches Zitnik Lab, Apache-2.0, skill dir confirmed, no OSV advisories

How to install

This skill calls ToolUniverse tools, so the ToolUniverse MCP server must be installed first (see the ToolUniverse page). Simplest registration:

claude mcp add --transport stdio tooluniverse -- uvx tooluniverse

Then add the skills:

Claude Code — install the whole skill collection (the skill resolves as tooluniverse-small-molecule-discovery):
```
npx skills add mims-harvard/ToolUniverse
```
Manual / other agents — copy just this skill directory into your skills folder:
```
git clone https://github.com/mims-harvard/ToolUniverse
cp -r ToolUniverse/skills/tooluniverse-small-molecule-discovery ~/.claude/skills/
```
(replace ~/.claude/skills/ with your agent’s skills directory if you are not using Claude Code/Desktop.)

The skill sets disable-model-invocation: true upstream, so invoke it explicitly (e.g. ask Claude to “use the small-molecule-discovery skill”) rather than relying on automatic dispatch.

What it does

Runs a six-phase compound-research workflow:

Compound identification — resolve chemical identity via PubChem CID, ChEMBL ID, and canonical SMILES.
Structure-based search — find analogs through similarity and substructure queries.
Bioactivity and binding — retrieve potency data (IC50, Ki) from ChEMBL and BindingDB (with a ChEMBL fallback).
Drug-likeness and ADMET — physicochemical rules and drug-likeness via SwissADME; ML-based toxicity, bioavailability, BBB penetrance, and CYP interactions via ADMET-AI.
Target prediction — infer likely protein targets for novel compounds (SwissTargetPrediction).
Commercial availability — check sourcing via eMolecules and Enamine catalogs (returns URLs when APIs are unavailable).

Primary use cases: hit characterization, analog scouting, ADMET triage, target deconvolution, and procurement of purchasable compounds.

Notes

It is a reasoning layer over ToolUniverse; without the MCP server registered, the tool calls fail. Overlaps but is distinct from the Chemical Safety skill (which centers hazard/toxicology) — this one centers identification, activity, and sourcing. Complements the standalone ChEMBL, PubChem, and BindingDB entries. ToolUniverse ships ~68 such skills; other workflows are catalogued separately.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.