BioChemAIgent: An AI-driven Protein Modeling and Docking Framework for Structure-Based Drug Discovery

Computational and AI-based methods have advanced drug discovery, yet most remain task-specific and require substantial expert integration. We introduce BioChemAIgent, an agentic framework that integrates state-of-the-art AI models with established computational chemistry tools to enable end-to-end small molecule analysis, protein modeling, and molecular docking, with a chatbot interface accessible online. To foster community engagement and extensibility, we established an openly accessible user interface alongside a centralized registry that consolidates resources for developing and integrating next-generation tools in drug discovery and structural biology.

Use BioChemAIgent

Online user interface

The online user interface is availabe at https://bcai.ims.bio/

Local Installation and Running BioChemAIgent via Docker

Download external MCP servers

cd mcp_external
git clone https://github.com/Augmented-Nature/PDB-MCP-Server.git
git clone https://github.com/Augmented-Nature/ChEMBL-MCP-Server.git

Build the image and run the conainer

docker build -t bcai-image .
docker run --rm -it -p 8501:8501 -v "$PWD":/workspace bcai-image bash

Use of external softwares: Instructions to install and run external softwares is provided in the software directory ./software

User access tokens

• To use cloud-based LLM models, three options are availabe: OpenAI [https://openai.com/api/], Ollama [https://ollama.com/], OpenRouter [https://openrouter.ai/]. These access tokens should be added to the env file.
• The use of ESM3 [https://github.com/evolutionaryscale/esm] and AlphaFold3 [https://github.com/google-deepmind/alphafold3] requires user access tokens. These access tokens should be added to the env file.

Contributions

This repository provides a community-oriented registry specialized for the development of agentic systems for drug discovery and structural biology.

• Domain experts can bring forward ideas to be implemented;
• Developers can propose or refine MCP servers to broaden and improve the agent’s functionality.

1. Create a new issue
2. Select the template
   • Add feature to suggest new functionalities to be added,
   • Add MCP server to suggest an existing MCP server.
3. Make a description of your suggestion and publish the issue.