Overview
CRISPR is no longer a graduate-only technique. With a good model organism, a validated protocol, and AI assistance, a motivated DIY biologist can run a clean knockout or knock-in experiment. This course walks you through the design, simulation, and wet-lab execution of a CRISPR edit while using Claude Code as your pipeline shell and Claude Cowork as a shared workspace for specialist agents.
How Claude Code and Claude Cowork fit
- Claude Code runs in your terminal. It drives primer design scripts, runs guide-RNA scorers, manages your Benchling or SnapGene files, and keeps a structured lab notebook in git.
- Claude Cowork is the collaborative workspace. You invite three specialists — a Safety Reviewer agent, a Guide-RNA Designer agent, and a Bioinformatics Analyst agent — into a shared project. They read the same files, comment on each other's work, and escalate to you when a decision is needed.
Who it's for
- Graduates of the DIY genetic engineering course
- Researchers prototyping edits in a personal or small-team lab
- Biotech-curious engineers who want to see end-to-end AI-assisted CRISPR
What you'll build
- A guide-RNA design and off-target analysis pipeline orchestrated by Claude Code
- A Claude Cowork project with Safety, Design, and Analysis agents collaborating on the edit
- A completed knockout or tagged knock-in in a model organism (yeast, worm, or K-12 E. coli derivative), validated by sequencing
Prerequisites
- Completed DIY genetic engineering course or equivalent bench skill
- Completed personal genome bioinformatics or comfort with command-line genomics
- A non-pathogenic model organism and an institutionally-appropriate project scope
Tools and setup
- Pick a target locus with a clear, observable phenotype
- Install Claude Code and a guide-RNA scoring toolchain
- Create a Claude Cowork project and add the three specialist agents
Modules
Module 1: Design with the Cowork team
You will draft guides with the Designer agent, stress-test them with the Safety Reviewer, and get off-target predictions from the Bioinformatics Analyst — all in a shared Claude Cowork project.
Module 2: Orchestrate the wet lab with Claude Code
You will drive ordering, protocol prep, and lab-notebook commits from Claude Code, so every experiment is reproducible and diffable.
Module 3: Edit, screen, and validate
You will deliver the edit, screen colonies or clones, and sequence-verify the outcome, with the Cowork agents helping interpret results.
Deliverable
A complete CRISPR project folder — guides, safety review, protocols, gel images, and sequencing confirmation — reproducible by another team using the same Claude Code and Cowork setup.
Common mistakes
- Picking a target without a clean readout and being unable to tell if the edit worked
- Skipping off-target analysis because the first guide "looked good"
- Running experiments the Safety Reviewer agent flagged without resolving the concern
Next steps
Extend into base editing, prime editing, or multiplex knockouts, and wire the Cowork project to your lab management system.