🧬 nf-core Agent for Claude Code

"Automate complex bioinformatics pipelines with a single natural language command."

This project introduces a custom slash command (/nf) to Anthropic's Claude Code, transforming it into a fully-fledged bioinformatics AI assistant. With a single prompt, the agent handles everything from environment validation and data acquisition to dynamic samplesheet generation, parameter tuning, and real-time error recovery for any nf-core pipeline (e.g., RNA-seq, WGS, Protein Structure Prediction).

✨ Key Features

1. 🗣️ Natural Language Pipeline Recommendation: Just describe your goal (e.g., "I want to find cancer mutations in my patient data"). The agent queries the live nf-core registry and recommends the most suitable pipeline (like nf-core/sarek).
2. 🛡️ Pre-flight Environment Validation: Before executing anything, it checks your Docker daemon, Nextflow version, Java runtime, and disk space to prevent runtime crashes.
3. 📥 Public Data (GEO/SRA) Auto-Acquisition: Provide a public accession ID (like GSE110004), and the agent will automatically download the required raw data.
4. 🧠 Universal Dynamic Samplesheet Engine: The agent fetches the specific schema_input.json for your chosen pipeline on the fly. Whether it requires FASTQ files with strandedness (genomics) or FASTA files (proteomics), it interactively builds a flawless samplesheet.csv.
5. 💡 Domain Expert Guide: During parameter setup, the agent scrapes and presents crucial biological tips and resource warnings specific to that pipeline (e.g., "Warning: AlphaFold DB requires 2.5TB of storage. Consider using ColabFold mode instead.").
6. 🚑 Real-time Monitoring & Auto-Recovery: If the pipeline fails (e.g., an OOM Exit 137 error), the agent steps in, explains the failure in plain English, and suggests a resume command with optimized parameters (like -max_memory).

🚀 Installation

You can install and inject the agent into any working directory with a single command via NPM.

Prerequisites

• Claude Code installed and authenticated
• Nextflow (v23.04+)
• Docker or Singularity
• Node.js & NPM (for installation)
• Python 3.8+

Quick Setup

Run the following command in the directory where your data lives (or where you want to run your analysis):

npx claude-nf-core-agent

This command will automatically create the .claude/commands/nf.md instruction file and copy the necessary backend Python scripts into your current directory without overwriting your existing data.

🎯 Usage

Once installed, launch Claude Code in that directory:

Bash

claude

Then, use the /nf command followed by your natural language request.

Example 1: Local RNA-seq Analysis

Bash

> /nf I want to analyze the gene expression levels of the raw data in my ./patients_fastq folder

Example 2: Public Data Re-analysis (WGS)

Bash

> /nf Download GSE110004 and find somatic variants

Example 3: Protein Structure Prediction

Bash

> /nf Predict the 3D structures for the protein sequences in my ./fasta_files directory

📂 Architecture

When installed, the package provisions the following structure:

• .claude/commands/nf.md: The master AI prompt defining the agent's workflow and behavior.
• scripts/: Executable automation scripts driven by the agent.
  • search_pipeline.py: Live nf-core API querying.
  • check_environment.py: System readiness validation.
  • generate_samplesheet.py: The dynamic, universal schema parser and generator.
  • run_nextflow.py: Interactive parameter tuning and deployment.
  • monitor_nextflow.py: Real-time log tailing and error diagnostics.