Skip to content

OpenCode

GitHub code

OpenCode CLI for AI-assisted development directly inside workspace containers.

See Last tested for the latest validation status.

To enable OpenCode CLI, set the OPENCODE argument to YES in the compose.yaml file of your desired workspace (e.g., template_ws/docker/compose.yaml). After making these changes, rebuild the Docker image.

Notice - Subscription required: OpenCode is a third-party service. This module only installs the CLI. You still need to sign in with a supported provider such as GitHub Copilot, Anthropic, OpenAI, Google, or others supported by OpenCode.

Installation

The module installs OpenCode CLI using the official installer from https://opencode.ai/install.

The current installer adds the CLI binary under ~/.opencode/bin, which is already available on the default container PATH.

Usage

After building a workspace with OpenCode enabled, you can use the CLI inside the container:

# Start your workspace container
cd ~/ros2-essentials/template_ws/docker
docker compose up -d
docker exec -it ros2-template-ws bash

# Inside the container, verify OpenCode is installed
opencode --version

# Run OpenCode CLI
opencode

The first time you run OpenCode, it will prompt you to choose and authenticate a model provider.

Configuring Local Models

Follow the OpenCode config locations guide and edit ~/.config/opencode/opencode.json directly.

The default compose.yaml mounts ${HOME}/docker/.config/opencode to /home/user/.config/opencode and ${HOME}/docker/.local/share/opencode to /home/user/.local/share/opencode so the installed binary and provider credentials can persist across containers.

For a local llama.cpp endpoint, add a provider entry like this by downloading and running the setup helper script:

curl -fsSL https://raw.githubusercontent.com/j3soon/local-llm-notes/refs/heads/main/examples/basic-secure-api/scripts/setup_opencode.sh -o /tmp/setup_opencode.sh
chmod +x /tmp/setup_opencode.sh
/tmp/setup_opencode.sh

The script installs the necessary dependencies and sets the OPENCODE_MODEL environment variable to point to the local model. After running it, you can start OpenCode as usual and it will use the locally-installed model.

Privacy

Note that using cloud-hosted models may allow your code to be used for training. See the OpenCode privacy policy for details: privacy policy.

References