Ash has built-in support for these AI coding agents. Each agent can be configured via ash.yml with the default settings below.
Run ash discover to auto-detect installed agents, or configure them manually in ash.yml.
| Agent | Binary | Args | Model Flag | Session Flag | Message Flag | Yes Flag | Install |
|---|---|---|---|---|---|---|---|
| opencode | opencode | run | --model | --continue | npm i -g @anomalyco/opencode | ||
| claude-code | claude | --model | --continue | --msg | npm i -g @anthropic-ai/claude-code | ||
| aider | aider | --model | --restore-chat-history | --msg | --yes | pip install aider-chat | |
| codex | codex | run | --model | --yolo | npm i -g @openai/codex | ||
| gemini-cli | gemini | -p | --model | --continue | npm i -g @google/gemini-cli | ||
| kimi | kimi | --model | --continue | -p | --yolo | curl -fsSL https://code.kimi.com/kimi-code/install.sh | bash | |
| pi | pi | --model | -p | npm i -g @picoglobal/pi | |||
| goose | goose | run | --model | -y | brew install goose | ||
| qwen-code | qwen | run | --model | pip install qwen-code | |||
| amazon-q | q | --model | -p | npm i -g @aws/q | |||
| echo | echo | Built-in (always available) |
Any CLI-based agent can be configured manually via ash.yml. Here is an example:
agents:
my-tool:
type: local-cli
cmd: my-tool
args: ["run"]
model_flag: "--model"
session_flag: "--continue"
message_flag: "--msg"
stdin_prompt: false
yes_flag: "--yes"See the full language reference for detailed documentation, or visit the GitHub repository.