ody init

Synopsis

ody init

Flag	Alias	Type	Default	Description
`--backend`	`-b`	`string`	(prompt)	Agent backend to use (`claude`, `opencode`, `codex`)
`--maxIterations`	`-i`	`string`	`"0"`	Max loop iterations. `0` means unlimited
`--model`	`-m`	`string`	(prompt)	Model identifier for the backend
`--autoCommit`	`-c`	`boolean`	`false`	Commit after each completed task
`--agent`	`-a`	`string`	(prompt)	Agent profile/persona for the backend harness
`--notify`	`-n`	`string`	(prompt)	Notification preference (`false`, `all`, `individual`)
`--dry-run`		`boolean`	`false`	Print the configuration without saving it

When run without flags, ody init walks you through an interactive setup:

Backend selection -- An autocomplete prompt lists backends detected on your system. Only backends with their CLI tool installed and on $PATH are shown.
Model selection -- Optional text input to specify a model for the backend. Leave blank to use the backend's default.
Agent profile -- Optional text input for the agent persona. Leave blank for the default (build).
Validation commands -- Prompts whether you want to add shell commands the agent should run to validate its work (e.g., bun lint, bun test). You can add multiple commands one at a time.
Claude permissions -- If the selected backend is claude, asks whether to skip Claude Code's permission checks.
Notifications -- Select when OS notifications should be sent: disabled, on full completion, or after each iteration.

The resulting configuration is saved to .ody/ody.json. If the .ody/ directory does not exist, it is created.

Interactive setup:

ody init

Non-interactive with all flags:

ody init --backend claude --maxIterations 5 --model claude-sonnet-4-20250514 --autoCommit

Preview what would be saved:

ody init --backend opencode --dry-run

File	Description
`.ody/ody.json`	Project configuration

See the Configuration guide for details on all config keys.