tama brew

tama brew compiles your tama project into a lean, production-ready Docker image. It runs tama lint first, installs all skill dependencies, and packages everything with the tamar runtime binary.

Usage

tama brew
tama brew --tag my-project:v1.2.0

What it does

1. Lint — runs tama lint and exits if any checks fail
2. Build base — starts from a distroless base image
3. Install dependencies — runs apt-get install and uv pip install for all skill depends:
4. Copy project files — copies agents/, skills/, tama.toml
5. Bundle binary — includes the tamar runtime binary
6. Tag image — tags as <project-name>:latest (or custom tag)

Output

A Docker image that runs tamar as the entrypoint:

docker run -e ANTHROPIC_API_KEY=sk-... \
           -e TAMA_MODEL_THINKER=anthropic:claude-opus-4-6 \
           my-project:latest \
           "research fusion energy trends"

Image size

The output image is typically ~8–15MB because it uses a distroless base (no shell, no apt, no unnecessary tools). This means fast cold starts and a minimal attack surface.

Configuration

Environment variables are not baked into the image. Pass them at runtime:

docker run \
  -e ANTHROPIC_API_KEY=... \
  -e TAMA_MODEL_THINKER=... \
  my-project:latest \
  "task input"

Custom tag

tama brew --tag my-project:v2.0.0
tama brew --tag registry.example.com/my-project:latest

Pushing to a registry

tama brew --tag registry.example.com/my-project:latest
docker push registry.example.com/my-project:latest