iterion

Delegation

For tasks that need full tool access (file editing, shell commands, git operations), you can delegate agent execution to an external CLI agent instead of making direct LLM API calls:

agent implementer:
  backend: "claude_code"          # recommended (codex is supported but discouraged)
  input: plan_schema
  output: result_schema
  system: implementation_prompt
  tools: [read_file, write_file, run_command, git_diff]

Backend	Status	What it does
`claude_code`	recommended	Runs the `claude` CLI as a subprocess with full tool access
`claw` (default)	recommended for read-only / judges	In-process multi-provider LLM client (Anthropic, OpenAI, …) — use with `model: "openai/gpt-5.4-mini"` etc.
`codex`	discouraged	Runs the `codex` CLI as a subprocess. Cannot configure its tool set, tends to fill its own context window, and has weaker iterion integration. The compiler emits a `C030` warning per node. Kept for compatibility — prefer `claude_code` or `claw`+OpenAI in new workflows.

💡 claude_code works with your Claude subscription (Pro/Max/Team/Enterprise) — no separate API key required. claw calls provider APIs directly and needs the corresponding API key (ANTHROPIC_API_KEY, OPENAI_API_KEY, …).

Delegation is useful for agents that need to act on the codebase (write files, run tests, execute commands). For agents that only need to think (review, judge, plan), use model: directly — it’s lighter weight and faster.

You can mix both in the same workflow. A common pattern is using model: for reviewers and judges, and backend: for implementers:

agent reviewer:
  model: "claude-sonnet-4-20250514"    # Direct API call — fast, read-only
  readonly: true

agent implementer:
  backend: "claude_code"              # Full agent — can edit files
  tools: [read_file, write_file, patch, run_command]

This site is open source. Improve this page.