---
title: RDKit Toolkit
description: Run RDKit descriptors, fingerprints, Lipinski and PAINS filters and Murcko scaffolds — in the sandbox up to 200 molecules, or as an rdkit_full job.
tags:
  - rdkit
---

RDKit turns SMILES into molecular descriptors, fingerprints, and drug-likeness filters, the fast, deterministic groundwork that prepares and interprets the inputs and outputs of heavier models. Reach for it whenever you need Lipinski or PAINS flags, similarity fingerprints, or format conversions without invoking an ML model. It complements synthesis planning in [AiZynthFinder](/docs/aizynthfinder) by profiling the molecules you route. See [what you can achieve](/docs/capabilities#chemistry) for the full set.

**RDKit** is the most widely used open-source toolkit for cheminformatics, developed by Greg Landrum with extensive community contributions. It provides a fast C++ backend with a versatile Python API, supporting molecular parsing, fingerprinting, descriptor calculation, substructure searching, and 3D conformer generation.

## Citations & Resources

> Landrum G, et al. RDKit: Open-source cheminformatics software. [https://www.rdkit.org](https://www.rdkit.org)

> **Documentation**: [https://www.rdkit.org/docs/](https://www.rdkit.org/docs/)

> **GitHub**: [https://github.com/rdkit/rdkit](https://github.com/rdkit/rdkit)

## Key capabilities

- **Molecular I/O**: Read/write SMILES, InChI, MOL, SDF, and PDB formats
- **Descriptors**: Calculate molecular properties for QSAR modeling
- **Fingerprints**: Morgan, RDKit, Atom Pair, and MACCS fingerprints
- **Database cartridge**: PostgreSQL integration for substructure/similarity searches
- **Version**: `rdkit>=2023.9.1` (pip)

RDKit on the platform covers fingerprinting, similarity, conformer generation, and cross-format conversions. Use the pages linked above for task-focused guides and copy-pasteable code.

## Where RDKit runs

RDKit runs on the platform's servers, not in your browser and not on a GPU.

- **In your chat's sandbox.** Descriptors, fingerprints, similarity, substructure search, standardization, drug-likeness filters and Murcko scaffolds on up to 200 molecules (100 for similarity) run inside your chat's [sandbox](/docs/sandbox), a container on AWS ECS Fargate that belongs to that chat alone. The result card completes in the same turn.
- **As a CPU job.** 3D conformers, MCS, diverse picking, CSV or SDF exports, and larger batches run as an `rdkit_full` job on a CPU worker. If the chat has no active sandbox, the lightweight operations fall back to this job too.
- **In your own code.** RDKit is preinstalled in the sandbox, so you can also ask Ivy to run your own RDKit Python against files in your workspace.

The 2D structure previews drawn in chat use RDKit.js in the browser, for display only; no analysis runs there.

## Chat Usage

### Example Prompts

- "Calculate RDKit descriptors for aspirin"
- "Run RDKit similarity analysis on my compound set"
- "Compute molecular fingerprints for these 10 SMILES"

### Workflow Chaining

- **Feeds into**: Compound Passport (molecular properties section), ADMET screening
- **Accepts from**: PubChem/ZINC/ChEMBL (SMILES), REINVENT/GenMol (generated molecules)

## FAQ

### How do I check which RDKit version I have?

  Ask Ivy to run `python -c "import rdkit; print(rdkit.__version__)"` in your chat's sandbox.

### What input formats are supported?

  SMILES, SDF, MOL, and PDB are supported for parsing, sanitization, and conversion.

## Read next