---
title: "AiChemy: Next"
description: "AiChemy is a multi-agent assistant for cross-disciplinary research that combines external MCP servers including OpenTargets, PubChem, and PubMed with Databricks-managed chemical libraries. It can identify disease targets and drug candidates, retrieve chemical and pharmacokinetic properties, assess safety and toxicity, and attach findings to verifiable data sources. Its guided workflow covers disease-to-target-to-drug discovery and literature validation, while a vector search over a 250,000-molecule ZINC15 index uses 1024-bit ECFP embeddings to find compounds similar to Elacestrant. To build a custom supervisor, users connect public MCP servers, transform structured drug data into a Genie space, and index unstructured chemical data for similarity search. They can assemble it with no-code Agent Bricks or deploy a Langgraph supervisor with Lakebase, MLflow AgentServer, and a React interface; invocations are logged and traced through MLflow using OpenTelemetry standards."
---

# AiChemy: Next

[Databricks](https://yomu.fyi/company/databricks) · Yen Low, Sean Zhang · Sep 18, 2026

**Type:** Tutorial

## Summary

AiChemy is a multi-agent assistant for cross-disciplinary research that combines external MCP servers including OpenTargets, PubChem, and PubMed with Databricks-managed chemical libraries. It can identify disease targets and drug candidates, retrieve chemical and pharmacokinetic properties, assess safety and toxicity, and attach findings to verifiable data sources. Its guided workflow covers disease-to-target-to-drug discovery and literature validation, while a vector search over a 250,000-molecule ZINC15 index uses 1024-bit ECFP embeddings to find compounds similar to Elacestrant. To build a custom supervisor, users connect public MCP servers, transform structured drug data into a Genie space, and index unstructured chemical data for similarity search. They can assemble it with no-code Agent Bricks or deploy a Langgraph supervisor with Lakebase, MLflow AgentServer, and a React interface; invocations are logged and traced through MLflow using OpenTelemetry standards.

## Context

The post describes AiChemy as a research-focused multi-agent supervisor that combines public biomedical and chemical MCP servers with structured and unstructured proprietary data managed on Databricks. It presents drug-discovery workflows, chemical-similarity lead generation, deployment choices, and observability features.

## Approach / What changed

The implementation connects OpenTargets, PubMed, and PubChem through Unity Catalog MCP connections, exposes structured drug data through a Genie space, and creates a vector index for molecular fingerprints in a chemical library. A no-code Agent Bricks supervisor can be deployed as a REST endpoint, while an advanced Langgraph implementation can add Skills and agentic memory through Lakebase, then deploy an MLflow AgentServer and React UI through Databricks Apps. MLflow experiments and OpenTelemetry-compatible traces support evaluation and monitoring.

## Takeaways

- AiChemy combines OpenTargets, PubMed, and PubChem with Databricks Genie and AI Search so external and proprietary knowledge bases can be analyzed together.
- The ZINC15 lead-generation workflow searches a 250,000-molecule index using Elacestrant's 1024-bit ECFP4 embedding to retrieve structurally similar compounds.
- The no-code supervisor deploys quickly through Agent Bricks, while the notebook-based Langgraph option supports Skills, agentic memory, Lakebase, and a custom React interface.

## FAQ

### How does AiChemy find chemically similar lead compounds?

It uses Databricks AI Search to query a 250,000-molecule ZINC15 index. Elacestrant's 1024-bit ECFP molecular embedding is used as the query vector to retrieve similar embeddings.

### What data sources are connected to the multi-agent supervisor?

The supervisor can connect to OpenTargets, PubMed, and PubChem through external MCP servers. It can also use a structured drug library through Genie and an unstructured chemical library with molecular fingerprint embeddings through AI Search.

### How are deployed agent invocations monitored?

Every invocation is logged and traced to a Databricks MLflow experiment using OpenTelemetry standards. Traces capture the execution graph, tool calls, retrieved documents, latency, and token usage.

**Tags:** [AI Agents](https://yomu.fyi/topic/ai-agents), [Databricks](https://yomu.fyi/topic/databricks), [LangGraph](https://yomu.fyi/topic/langgraph), [MCP](https://yomu.fyi/topic/mcp), [MLflow](https://yomu.fyi/topic/mlflow)

- Source: [Databricks](https://www.databricks.com/blog/aichemy-next-generation-agent-mcp-skills-and-custom-data-drug-discovery)
- Source URL: https://www.databricks.com/blog/aichemy-next-generation-agent-mcp-skills-and-custom-data-drug-discovery
- Ingested by Yomu: 2026-09-17T14:06:27.370Z

[Read original post](https://www.databricks.com/blog/aichemy-next-generation-agent-mcp-skills-and-custom-data-drug-discovery)