Example coding-agent commands

These are examples, not an exclusive list of coding agents. You can bring another agent through a compatible local command that meets the command-template requirements.

Start with one of these examples when you configure a coding agent. The tested CLI version and operating system are listed below each provider. Variants without a successful listener launch are labeled unverified.

claude --model {Model:opus|sonnet|haiku} --effort {Effort:high|medium|low|max} --permission-mode acceptEdits --verbose --output-format stream-json "{alterac-ai-message}"

Tested with Claude Code 2.1.280 on macOS.

fable is an optional model value for accounts with access. It is not included in this tested example and remains unverified. See the Claude Code CLI reference (external site) for provider options.

codex --model {Model:gpt-6.1-sol|gpt-6-astra} --sandbox workspace-write --ask-for-approval on-request -c 'approvals_reviewer="auto_review"' -c 'sandbox_workspace_write.network_access=true' -c 'model_reasoning_effort="{Effort:xhigh|high|medium|low|max}"' -c 'model_reasoning_summary="concise"' -c 'show_raw_agent_reasoning=false' -c 'model_verbosity="low"' --search "{alterac-ai-message}"

Tested with Codex 0.159.2 on macOS.

Review the sandbox, approval, network, and search settings for your repository. See the Codex CLI reference (external site) for provider options.

agent --model auto "{alterac-ai-message}"

Tested with Cursor CLI 2026.10.01-e373342 on macOS.

Auto delegates model selection to Cursor. The terminal remains interactive, allowing questions and follow-up instructions. See the Cursor CLI overview (external site) for installation and authentication.

This optional Cursor CLI template has not been verified through a successful listener launch:

agent --model claude-{Model:opus|sonnet}-5-5-{Effort:medium|high|low|xhigh|max} "{alterac-ai-message}"

Selecting sonnet and high resolves to claude-sonnet-5-5-high. All model/effort combinations above appeared in the installed model list; Haiku was absent and is excluded.

Named-model execution was blocked on the tested Free plan, which allowed only Auto. This template requires a plan permitting named models and its execution remains unverified. See the Cursor CLI parameters reference (external site) for provider options.