Best for
- Researchers and developers integrating chemical database queries into LLM-driven tools.
- Those preferring a public hosted endpoint or self-hosting their own MCP server.
This is a feature-rich and well-documented server for users wanting to integrate PubChem data into AI workflows. It offers a comprehensive set of tools and resources with easy setup. Note that it is not an official PubChem project, and the public endpoint is community-hosted.
PubChem MCP Server is a community-maintained server that enables AI assistants to query the PubChem chemical database. It provides tools for searching compounds by name, structure, etc., fetching detailed information, images, 3D structures, external cross-references, safety data, bioactivity, interactions, and entity summaries. It also exposes resources for direct access to compound and assay data. It supports Streamable HTTP and can be used via a public hosted instance or self-hosted locally over stdio or HTTP. No API keys are required, and all tools are read-only.
Install Bun v1.3.0 or higher (or Node.js v24+). Then add a configuration to your MCP client (e.g., claude_desktop_config.json) with type 'stdio', command 'bunx', and args ['@cyanheads/pubchem-mcp-server@latest']. No API keys are required.
{
"mcpServers": {
"pubchem-mcp-server": {
"type": "stdio",
"command": "bunx",
"args": [
"@cyanheads/pubchem-mcp-server@latest"
],
"env": {
"MCP_TRANSPORT_TYPE": "stdio"
}
}
}
}