Installation

Table of contents

1. Prerequisites
2. Step 1 — Create the Python Environment
3. Step 2 — Install Claude Code CLI
4. Step 3 — Verify the Installation
5. Installing MCPs
   1. Docker MCPs (local build recommended)
   2. Auto-install (pulls from registry)
   3. Python MCPs (no Docker needed)
   4. Verify Installed MCPs

---

Prerequisites

The following tools must be installed on your system:

Tool	Purpose	Install Guide
Python 3.10+	Core runtime	python.org
Conda/Mamba	Environment management	miniforge
Node.js / npm	Claude Code CLI	nodejs.org
Docker (with GPU support)	Containerized MCP servers	docs.docker.com
NVIDIA drivers + nvidia-container-toolkit	GPU access in Docker	NVIDIA Container Toolkit

Verify your setup:

python --version       # >= 3.10
conda --version        # or mamba --version
npm --version
docker --version
nvidia-smi             # GPU available
docker run --rm --gpus all nvidia/cuda:12.1.0-base-ubuntu22.04 nvidia-smi  # GPU in Docker

---

Step 1 — Create the Python Environment

mamba env create -f environment.yml
mamba activate protein-mcp
pip install -r requirements.txt
pip install -e .

Step 2 — Install Claude Code CLI

npm install -g @anthropic-ai/claude-code

Step 3 — Verify the Installation

pmcp avail     # List all available MCPs
pskill avail   # List all available workflow skills
claude --version

---

Installing MCPs

MCPs come in two runtime types:

Type	MCPs	Install method
Python (local venv)	msa_mcp, alphafold2_mcp, mmseqs2_mcp, …	quick_setup.sh creates a local env/ venv
Docker (GPU container)	esm_mcp, prottrans_mcp, plmc_mcp, ev_onehot_mcp, bindcraft_mcp, boltzgen_mcp	Docker image build or pull

Docker MCPs (local build recommended)

Local builds are faster than pulling from the registry:

cd tool-mcps/esm_mcp && docker build -t esm_mcp:latest . && cd ../..
cd tool-mcps/prottrans_mcp && docker build -t prottrans_mcp:latest . && cd ../..
cd tool-mcps/plmc_mcp && docker build -t plmc_mcp:latest . && cd ../..
cd tool-mcps/ev_onehot_mcp && docker build -t ev_onehot_mcp:latest . && cd ../..
cd tool-mcps/bindcraft_mcp && docker build -t bindcraft_mcp:latest . && cd ../..
cd tool-mcps/boltzgen_mcp && docker build -t boltzgen_mcp:latest . && cd ../..

Then register with Claude Code:

# pmcp install detects the local image and skips pulling from registry
pmcp install esm_mcp
pmcp install prottrans_mcp
pmcp install plmc_mcp
pmcp install ev_onehot_mcp
pmcp install bindcraft_mcp
pmcp install boltzgen_mcp

Auto-install (pulls from registry)

pmcp install esm_mcp        # Pulls ghcr.io/macromnex/esm_mcp:latest

Python MCPs (no Docker needed)

pmcp install msa_mcp         # Runs quick_setup.sh, creates local venv

Verify Installed MCPs

pmcp status                  # Shows installed/registered status
claude mcp list              # Health-check all registered MCPs