ExperimentPrototype
Multi-Agent Monorepo Packaging Strategy
One repository, many agents, independent deploys — path-filtered pipelines over a shared base image, so editing one agent doesn't rebuild the rest.
What I'm testing
Whether a shared base image with per-agent thin layers beats fully independent images once you have more than about five agents.
Finding so far
Build time improves substantially. The tradeoff is coupling: a base image change rebuilds everything, so the base has to be genuinely stable before this pays off.
Open question
Where the crossover point sits. Below roughly four agents, independent images seem simpler and the build time saved isn't worth the coupling.
Stack
- Cloud Build
- Docker
- GitHub Actions