PubChem MCP Server
Provides comprehensive access to PubChem's chemical information database via the PubChem PUG REST API.
@cyanheads/pubchem-mcp-server
MCP server for the PubChem chemical database. Search compounds, fetch properties, safety data, bioactivity, cross-references, and entity summaries. STDIO & Streamable HTTP
8 Tools
Public Hosted Server: https://pubchem.caseyjhand.com/mcp
Tools
Eight tools for querying PubChem's chemical information database:
| Tool Name | Description |
|---|---|
pubchem_search_compounds | Search for compounds by name, SMILES, InChIKey, formula, substructure, superstructure, or 2D similarity. |
pubchem_get_compound_details | Get physicochemical properties, descriptions, synonyms, drug-likeness, and classification for compounds by CID. |
pubchem_get_compound_image | Fetch a 2D structure diagram (PNG) for a compound by CID. |
pubchem_get_compound_safety | Get GHS hazard classification and safety data for a compound. |
pubchem_get_compound_xrefs | Get external database cross-references (PubMed, patents, genes, proteins, etc.). |
pubchem_get_bioactivity | Get a compound's bioactivity profile: assay results, targets, and activity values. |
pubchem_search_assays | Find bioassays by biological target (gene symbol, protein, Gene ID, UniProt accession). |
pubchem_get_summary | Get summaries for PubChem entities: assays, genes, proteins, taxonomy. |
pubchem_search_compounds
Search PubChem for chemical compounds across five search modes.
- Identifier lookup — resolve compound names, SMILES, or InChIKeys to CIDs (batch up to 25)
- Formula search — find compounds by molecular formula in Hill notation
- Substructure/superstructure — find compounds containing or contained within a query structure
- 2D similarity — find structurally similar compounds by Tanimoto similarity (configurable threshold)
- Optionally hydrate results with properties to avoid a follow-up details call
pubchem_get_compound_details
Get detailed compound information by CID.
- Batches up to 100 CIDs in a single request
- 27 available properties: molecular weight, SMILES, InChIKey, XLogP, TPSA, complexity, stereo counts, and more
- Optionally includes textual descriptions (pharmacology, mechanism, therapeutic use) from PUG View
- Optionally includes all known synonyms (trade names, systematic names, registry numbers)
- Optionally computes drug-likeness assessment (Lipinski Rule of Five + Veber rules) from fetched properties
- Optionally fetches pharmacological classification (FDA classes, mechanisms of action, MeSH classes, ATC codes)
pubchem_get_bioactivity
Get a compound's bioactivity profile from PubChem BioAssay.
- Returns assay outcomes (Active/Inactive/Inconclusive), target info (protein accessions, NCBI Gene IDs), and quantitative values (IC50, EC50, Ki)
- Filter by outcome to focus on active results
- Caps at 100 results per request (well-studied compounds may have thousands)
pubchem_get_summary
Get descriptive summaries for four PubChem entity types.
- Assays (AID), genes (Gene ID), proteins (UniProt accession), taxonomy (Tax ID)
- Up to 10 entities per call
- Type-specific field extraction for clean, structured output
Features
Built on @cyanheads/mcp-ts-core:
- Declarative tool definitions — single file per tool, framework handles registration and validation
- Unified error handling across all tools
- Pluggable auth (
none,jwt,oauth) - Swappable storage backends:
in-memory,filesystem,Supabase,Cloudflare KV/R2/D1 - Structured logging with optional OpenTelemetry tracing
- Runs locally (stdio/HTTP) or containerized via Docker
PubChem-specific:
- Rate-limited client for PUG REST and PUG View APIs (5 req/s with automatic queuing)
- Retry with exponential backoff on 5xx errors and network failures
- All tools are read-only and idempotent — no API keys required
Getting Started
Public Hosted Instance
A public instance is available at https://pubchem.caseyjhand.com/mcp — no installation required. Point any MCP client at it via Streamable HTTP:
{
"mcpServers": {
"pubchem": {
"type": "streamable-http",
"url": "https://pubchem.caseyjhand.com/mcp"
}
}
}
Self-Hosted / Local
Add to your MCP client config (e.g., claude_desktop_config.json):
{
"mcpServers": {
"pubchem": {
"type": "stdio",
"command": "bunx",
"args": ["@cyanheads/pubchem-mcp-server@latest"],
"env": {
"MCP_TRANSPORT_TYPE": "stdio"
}
}
}
}
Prerequisites
- Bun v1.2.0 or higher (or Node.js v22+)
Installation
- Clone the repository:
git clone https://github.com/cyanheads/pubchem-mcp-server.git
- Navigate into the directory:
cd pubchem-mcp-server
- Install dependencies:
bun install
Configuration
No API keys are required — PubChem's API is freely accessible.
| Variable | Description | Default |
|---|---|---|
MCP_TRANSPORT_TYPE | Transport: stdio or http. | stdio |
MCP_HTTP_PORT | Port for HTTP server. | 3010 |
MCP_HTTP_HOST | Host for HTTP server. | localhost |
MCP_AUTH_MODE | Auth mode: none, jwt, or oauth. | none |
MCP_LOG_LEVEL | Log level (RFC 5424). | info |
STORAGE_PROVIDER_TYPE | Storage backend. | in-memory |
OTEL_ENABLED | Enable OpenTelemetry. | false |
Running the Server
Local Development
-
Build and run:
bun run rebuild bun run start:stdio # or start:http -
Run checks and tests:
bun run devcheck # Lints, formats, type-checks bun run test # Runs test suite
Docker
docker build -t pubchem-mcp-server .
docker run -p 3010:3010 pubchem-mcp-server
Project Structure
| Directory | Purpose |
|---|---|
src/mcp-server/tools/definitions/ | Tool definitions (*.tool.ts). |
src/services/pubchem/ | PubChem API client with rate limiting and response parsing. |
scripts/ | Build, clean, devcheck, and tree generation scripts. |
Development Guide
See CLAUDE.md for development guidelines and architectural rules. The short version:
- Handlers throw, framework catches — no
try/catchin tool logic - Use
ctx.logfor domain-specific logging - Register new tools in the
index.tsbarrel file
Contributing
Issues and pull requests are welcome. Run checks before submitting:
bun run devcheck
bun run test
License
Apache-2.0 — see LICENSE for details.
相关服务器
VictoriaMetrics
A server for writing and querying time series data using the VictoriaMetrics API.
FoodData Central
Access the USDA's FoodData Central database for comprehensive food and nutrient information.
Eka MCP Server
Access medical knowledge-bases and drug information from eka.care. Requires API credentials.
Redis
A server for interacting with Redis databases.
Engram MCP Server
Engram is a hosted MCP server that provides reliable memory for AI agents:
Teradata
A collection of tools for managing the platform, addressing data quality and reading and writing to Teradata Database.
AKShare One
Access Chinese stock market data, including historical prices, real-time quotes, news, and financial statements.
Google Directory by CData
A read-only MCP server for querying live Google Directory data using the CData JDBC Driver.
Simple PostgreSQL MCP Server
An MCP server for interacting with PostgreSQL databases using tools, resources, and prompts.
DeepMemory
DeepMemory MCP is a small Model Context Protocol (MCP) server that provides long-term memory storage for conversational agents.