# Deep MedChem

> Deep MedChem builds CHEESE, AI-powered search over the whole purchasable chemical space (built for 100T+ molecules; ~18T make-on-demand molecules live today, plus in-stock catalogues), searchable by 2D fingerprint (Morgan/ECFP4), 3D shape and electrostatic similarity. AI assistants and agents can use it through a hosted MCP server with no account or API key, through plugins for Claude Code and ChatGPT, or through the `deepmedchem` Python SDK.

Deep MedChem is making the whole purchasable chemical space searchable. The platform is built for more than 100 trillion molecules, so an AI agent can search trillions of compounds, pick candidates with list prices and vendor IDs, and order them: make-on-demand molecules are synthesized by the vendor in a few weeks for a low per-compound price, and in-stock catalogues ship immediately.

Searchable spaces (sizes from the live catalog, September 2026):

| Space | Molecules | Lead time | Reported synthesis success |
| --- | ---: | --- | --- |
| Synple eXplore | 9.5T | 3-4 weeks | >85% |
| Synple | 7.6T | 3-4 weeks | >85% |
| ChemInfinita | 794B | 5-8 weeks | 55-85% |
| Freedom Space 5 | 296B | 5-6 weeks | >80% |
| Enamine REAL | ~93B combinations | 3-4 weeks | >80% |
| VAST | 6.8B | 2-4 weeks | 85%+ |
| D2B SpaceM1 | 1.5B | 2-6 weeks | >85% |
| Onepot CORE 2.0 Lite | - | 5-7 business days | 60-70% |

Also: a 100-trillion-molecule DeepMedChem demo space and in-stock catalogues (for example Mcule amino acids) in the next-generation CHEESE preview. Full, current breakdown: https://github.com/Deep-MedChem/deepmedchem-python#deepmedchem-all-chemical-spaces

Key facts:

- MCP server: `https://mcp.deepmedchem.com/mcp` (streamable HTTP, no authentication). Tools: `catalog`, `similarity_search` (one SMILES, one database, up to 20 hits), `render_results`.
- Claude Code: `/plugin marketplace add Deep-MedChem/plugin`, then `/plugin install deepmedchem@deepmedchem`.
- Guest MCP covers most of the make-on-demand spaces above; call `catalog` for current IDs, sizes and prices. Ordering, quotes, substructure search, filters and in-stock catalogues need the Python SDK (`dmc order results.csv --get-quote`) or CHEESE.
- Similarity scores describe structural similarity, not biological activity or safety.

## For AI agents

- [Agent skill](https://www.deepmedchem.com/skill.md): how to connect to the MCP server (or plain HTTP) and the rules for running chemical-space searches
- [LLM install guide](https://www.deepmedchem.com/llms): set up DeepMedChem in Claude Code, Claude.ai, ChatGPT and other MCP clients
- [Plugin source](https://github.com/Deep-MedChem/plugin): Claude and ChatGPT plugin manifests, skill and privacy notes
- [ChatGPT plugin](https://chatgpt.com/plugins/plugin_asdk_app_6ace45caacf0819194fd9a0a63e74873): install DeepMedChem in ChatGPT

## Products

- [CHEESE](https://cheese.deepmedchem.com/): web app for sub-second similarity search across chemical spaces
- [Documentation](https://docs.deepmedchem.com/): CHEESE API v2, search, selections and durable runs ([llms.txt](https://docs.deepmedchem.com/llms.txt))
- [Python SDK and CLI](https://github.com/Deep-MedChem/deepmedchem-python): `pip install deepmedchem`, the `dmc` command (search, substructure, sampling, quotes and orders, Navigator optimization) and its agent skill
- [Next-generation CHEESE (preview)](https://cheese-new-dev.deepmedchem.com/): synthon-native search, including the 100T demo space and in-stock catalogues

## Company

- [Home](https://www.deepmedchem.com/): products, team and contact
- [News](https://www.deepmedchem.com/news): product releases and newsletters
- [Case studies](https://www.deepmedchem.com/cases): customer results
- [147x hit rate case study](https://www.deepmedchem.com/case-study/147x-hit-rate)

## Optional

- [Blog](https://deepmedchem.substack.com/)
- [Presentations](https://slides.deepmedchem.com/)
- [Privacy policy](https://cheese.deepmedchem.com/privacy-policy)
- [Terms of service](https://cheese.deepmedchem.com/terms-and-conditions)
- Contact: info@deepmedchem.com