Install¶

Pick one of the four paths below. All produce the same spatial-cluster binary speaking MCP over stdio.

Option 1 — Nix (recommended)¶

The flake provisions Go, GEOS, pkg-config and the static toolchain deterministically. This is what CI uses.

git clone https://github.com/kartoza/kartoza-mcp.git
cd kartoza-mcp
nix develop          # drops you into a shell with everything wired up
nix run .#build      # produces ./spatial-cluster-static

You can also build for Linux explicitly:

nix run .#build-all  # produces ./spatial-cluster-linux-amd64

Option 2 — `go install`¶

If you already have Go 1.25+ and GEOS on your PKG_CONFIG_PATH, you can pull the latest tagged release straight into your GOBIN:

go install github.com/kartoza/kartoza-mcp@latest

The binary is named spatial-cluster and runs on stdio out of the box.

Option 3 — Prebuilt static binary¶

Every GitHub release ships a static Linux binary built with musl. No shared libraries needed.

curl -fsSL https://github.com/kartoza/kartoza-mcp/releases/latest/download/spatial-cluster-linux-amd64 \
  -o spatial-cluster
chmod +x spatial-cluster
./spatial-cluster --help

See DevOps Deployment for distro-aware install patterns.

Option 4 — Container image¶

For Docker, Podman or Kubernetes deployments:

docker pull ghcr.io/kartoza/kartoza-mcp:latest
docker run --rm -i ghcr.io/kartoza/kartoza-mcp:latest < test.json

The image is FROM scratch plus the static binary — under 10 MB.

Verify the install¶

./spatial-cluster-static < test.json | head -c 200

You should see a GeoJSON FeatureCollection printed to stdout. If you do — you're ready for your first cluster.

GEOS at runtime

The static binary embeds GEOS — you do not need a system libgeos. If you build dynamically (e.g. go build outside the flake) you'll need GEOS available via pkg-config. See Developers Project structure.