Skillget

rdkit-mcp-server

mcp
Guvenlik Denetimi
Uyari
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 39 GitHub stars
Code Uyari
  • process.env — Environment variable access in .github/workflows/run-evals.yaml
  • fs module — File system access in .github/workflows/run-evals.yaml
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose
This MCP server bridges large language models with RDKit, enabling AI agents to execute chemistry and molecule-related functions using natural language commands.

Security Assessment
Overall risk: Low. The server does not request dangerous permissions, execute direct shell commands, or contain hardcoded secrets. The only flagged warnings involve environment variable and file system access, which are safely contained entirely within an automated GitHub Actions workflow (run-evals.yaml) rather than the core application code. The tool requires an OpenAI API key to use its CLI client, but this is properly handled via standard environment variables instead of being exposed in the code. It acts strictly as a local bridge without silently exfiltrating sensitive data.

Quality Assessment
The project demonstrates strong health and maintenance signals. It is actively maintained, with repository updates pushed as recently as today. It is fully licensed under the permissive MIT license, making it suitable for most development and commercial environments. Community trust is in its early stages but shows positive traction with 39 GitHub stars. Documentation is clear, thorough, and includes helpful usage examples alongside evaluation instructions.

Verdict
Safe to use.
SUMMARY

MCP server that enables language models to interact with RDKit through natural language

README.md

RDKit MCP Server: Agentic Access to RDKit for LLMs

RDKit MCP Server is an open-source MCP server that enables language models to interact with RDKit through natural language. The goal is to provide agent-level access to every function in RDKit 2025.3.1 without writing any code.

Features

  • Seamless Integration: Exposes RDKit functions via the Model Context Protocol (MCP).
  • Language Model Support: Connect any LLM that supports the MCP protocol.
  • CLI Client: Includes a command-line client powered by OpenAI for quick experimentation.

Table of Contents

Installation

Install the package:

pip install .

Usage

Start the Server

python run_server.py [--settings settings.yaml]

See settings.example.yaml for setting options

Once the server is running, any MCP-compliant LLM can connect. For example, see the Claude Desktop quickstart.

CLI Client

A CLI client is included for rapid prototyping with OpenAI:

export OPENAI_API_KEY="sk-proj-xxx"
python run_client.py

Available Tools

List all available RDKit tools exposed by the server:

python list_tools.py [--settings settings.yaml]

Evaluations

The evals directory contains a test suite for evaluating RDKit MCP tool outputs and agent responses using pydantic-evals.

Install Dependencies

pip install ".[evals]"

Start the MCP Server

In one terminal, start the server:

python run_server.py

Run Evaluations

In another terminal, run the evaluation suite:

python evals/run_evals.py

Options:

  • --verbose - Show detailed output including inputs and outputs
  • --filter <name> - Run only cases matching the name
  • --output-json results.json - Export results to JSON

Each test uses LLM-based evaluation (LLMJudge) to assess whether the agent correctly used the RDKit tools and produced accurate results.

Contributing

We welcome contributions, feature requests, and bug reports:

See CONTRIB.md for guidelines on how to get started.

Together, we can make RDKit accessible to a wider range of applications through natural language interfaces.

Yorumlar (0)

Sonuc bulunamadi