Background Agents

My favourite thing about Github Agents is that they run in their own isolated containers and can scale infinitely.

Until recently I’ve felt a bit stuck with Claude Code. On one hand its the best code generator. On the other, it doesn’t scale well beyond one instance and I have to wait for it to complete, which sometimes can be quite a while.

Now I’ve figured out a way to work on multiple tickets at a time by running multiple instances of claude code. This is something they talk about in the docs, albeit a rather brief section called ”Run parallel Claude Code sessions with Git worktrees”. But in most applications there are external dependencies (like a database) that are shared across worktrees.

Short of dockerising everything including Claude (which I might attempt next) I decided to create a new database in the postgres cluster for each worktree.

This keeps things nice and isolated so I can run migrations and make database changes whilst running multiple parallel claude code instances.

Then when I’m finished I can open a pull request (every worktree has its own branch) and remove the git worktree.

There are a few drawbacks with this method though:

• For one, the other resources the app connects to are still shared.
• I need to remember to delete the databases when I remove the worktree (although I have a script for this).
• It can take a while to create a new worktree as I need to install all the dependencies and compile each time.
• Claude could still do crazy things that affect my system.

Running Claude Code in isolated containers is really the gold standard here, I could potentially pre-compile the dependencies, have a docker container with everything already setup, have no shared dependencies, and not worry about Claude going off the rails on my machine.

I’ll have a play and see how far I get that way.