Parallel Coding Agents with Container Use and Git Worktree
Description
AI ENGINEER ROADMAP [ 🚀 learn AI Engineering in 2025 ]► https://zazencodes.com/NEWSLETTER [ 🍰 weekly video release email ]► https://zazencodes.com/newslett...
Notes
00:00 Introduction to Parallel Agents
- Parallel agents represent the future of generative coding
- Three workflows for safely developing with parallel agents in a single codebase
- Demonstrations use Claude Code but applicable to other coding agents (Cursor, Gemini)
- Workflows build in sophistication with container use as the most powerful approach
- 00:30: Container use is an open-source project from Dagger, created by Docker's founders
00:51 Workflow 1: Copy Folders Method
Overview
- Simplest approach: copy the project folder to separate directories for each agent
- Each agent works on a different feature implementation in parallel
Implementation
Comparison and Merging
05:24 Workflow 2: Git Worktree Method
Advantages Over Copy Folders
- Faster: creates branches and workspaces instantly without full folder duplication
- Uses Git best practices by creating separate branches for each agent
- Avoids copying entire file structures
Setup
- 05:40: Use
git worktree add -b [branch-name] [folder-name]to create agent workspaces - Automatically creates feature branches (agent A and agent B)
- Each agent works on their respective Git branch
Development Process
- 06:42: Navigate into each agent folder with full Git environment access
- 06:52: Still need to run
npm installin each workspace (downside) - Run dev servers on different ports (agent A on 8080, agent B on 8081)
- 07:30: Build different feature variations (e.g., click-through rate focus vs. aesthetics focus)
Merging Changes
11:58 Workflow 3: Container Use with Dagger
Introduction and Installation
- Most powerful approach for parallel agent development
- Built by creators of Docker, newly launched open-source project
- 12:07: Install via
brew install dagger-io/dagger/container-use - 12:15: Check version with
container-use version - 12:22: List environments with
container-use list, watch withcontainer-use watch
MCP Server Integration
Quick Start Demo
- 14:17: Create Python hello world script and update readme
- 14:29:
container-use environment createspawns containerized workspace - 15:00: Work happens in isolated container, files don't appear locally until checkout
- 15:31:
container-use checkout [environment-name]pulls changes to local directory - Creates corresponding Git branch to track all work
- 15:57: All changes are automatically committed with messages
- 16:14:
container-use mergebrings changes to main branch - 16:50:
container-use deleteremoves environment after merge
Real-World Multi-Agent Demo
- 17:39: Open same project in four different terminal panes
- Run Claude Code in parallel with
--allowed-toolsflag to skip confirmations - 18:19: Run three simultaneous redesign tasks:
- Redesign website with fantasy RPG style theme
- Redesign website with Barbie theme
- Redesign website with super minimal feel
- 18:35: Each task receives unique environment ID with separate containers
- All agents work in isolation without stepping on each other
Viewing and Comparing Results
- 19:14:
container-use diff [environment-name]shows all changes made - 19:31:
container-use checkout [environment-name]switches to each variant - Run
npm run devto preview each redesign locally - 20:35: Review different theme implementations and select preferred version
- Each environment has complete Git history with all commits
Deployment
Key Advantages
- 17:25: Requires Docker running on system
- Complete isolation: no dependency conflicts between parallel agents
- Automatic Git commits with descriptive messages
- Simplified workflow compared to git worktree
- No manual npm install needed per agent
- Branch creation and cleanup automated
23:30 Conclusion and Future Outlook
- Parallel agents represent the future of AI code generation
- Benefits demonstrated: quick feature iteration, version comparison, clean selection
- Container use workflow will only get easier with future updates
- Dagger created by Docker founders, positioned as the standard for parallel agent workflows
- Call to action: like video, subscribe, and try container-use implementation
Parallel Coding Agents with Container Use and Git Worktree
Interactive graph
On this page
Description
Notes
00:00 Introduction to Parallel Agents
00:51 Workflow 1: Copy Folders Method
Overview
Implementation
Comparison and Merging
05:24 Workflow 2: Git Worktree Method
Advantages Over Copy Folders
Setup
Development Process
Merging Changes
11:58 Workflow 3: Container Use with Dagger
Introduction and Installation
MCP Server Integration
Quick Start Demo
Real-World Multi-Agent Demo
Viewing and Comparing Results
Deployment
Key Advantages
23:30 Conclusion and Future Outlook