Skip to content

Discussion: Git Worktrees #1

@colinmollenhour

Description

@colinmollenhour

Hi Bryan, great project! I love how small it is. One issue that I'm thinking about is multiple agents conflicting with each other on the working copy. One solution for this could be to run multiple agents in different git worktrees. It might be as simple as some additional instructions in the directory like:

  • when starting a new feature, create a new worktree on a new branch (maybe some keyword like "major feature" or "worktree feature")
  • run a script like "pnpm run prepare" to perform any prep steps (e.g. creating a .env file, seeding a database, etc)
  • perform the work
  • commit changes
  • merge them into the main branch and resolve conflicts and run tests

Of course that can be done manually which is what I'm doing now, I'm wondering if with a workflow like this it would be more feasible to create agents that work in parallel as long as the task tree is well-designed. Maybe I'm overthinking it, but curious about your thoughts and experiences with using track-cli - I just recently discovered it so haven't actually used it yet. :D

I just came from working with "claude-flow" which I found to be a big disorganized monstrosity of dozens of features all slapped together which sounded nice, but in practice was horrible. The MCP servers consumed over 120k tokens, lol. But I liked the overall idea of being able to kick off larger tasks and having multiple agents working in parallel.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions