> ## Documentation Index
> Fetch the complete documentation index at: https://docs.airun.me/llms.txt
> Use this file to discover all available pages before exploring further.

# Tool, Effort & Profile Flags

> Select your runtime, control reasoning depth, and choose provider profiles

Flags for selecting the AI runtime, controlling reasoning effort, and choosing provider configurations.

## Runtime Selection

<ParamField path="--tool" type="flag">
  **Select AI runtime** by name

  ```bash theme={null}
  ai --tool cc task.md        # Claude Code
  ai --tool codex task.md     # Codex CLI
  ```

  If not specified, AI Runner auto-detects: tries Claude Code first, falls back to Codex CLI.
</ParamField>

<ParamField path="--cc" type="flag">
  **Claude Code** — Anthropic's AI coding CLI (default if installed)

  ```bash theme={null}
  ai --cc                     # Interactive Claude Code session
  ai --cc --aws --opus        # Claude Code with AWS Bedrock
  ai --cc task.md             # Execute script with Claude Code
  ```

  Shorthand for `--tool cc`. This is the default runtime when both are installed.
</ParamField>

<ParamField path="--codex" type="flag">
  **Codex CLI** — OpenAI's coding agent

  ```bash theme={null}
  ai --codex                  # Interactive Codex session
  ai --codex --high           # Codex with gpt-5.4
  ai --codex task.md          # Execute script with Codex
  ai --codex --ollama         # Codex with local Ollama
  ```

  Shorthand for `--tool codex`.

  **Authentication:** Codex uses browser login by default (`codex login`). Set `OPENAI_API_KEY` or `CODEX_API_KEY` for API key auth, or use `--apikey` to require it.

  **Install:**

  ```bash theme={null}
  npm install -g @openai/codex
  # Or on macOS:
  brew install --cask codex
  ```
</ParamField>

## Effort Level

<ParamField path="--effort" type="flag">
  **Reasoning effort** — control how deeply the model thinks

  ```bash theme={null}
  ai --effort low task.md       # Minimal reasoning (fast)
  ai --effort medium task.md    # Balanced (default)
  ai --effort high task.md      # Deep reasoning
  ai --effort max task.md       # Maximum reasoning
  ```

  Works with both runtimes. AI Runner maps to each tool's native format:

  | AI Runner | Claude Code                | Codex CLI                          |
  | --------- | -------------------------- | ---------------------------------- |
  | `low`     | `--effort low`             | `-c model_reasoning_effort=low`    |
  | `medium`  | `--effort medium`          | `-c model_reasoning_effort=medium` |
  | `high`    | `--effort high`            | `-c model_reasoning_effort=high`   |
  | `max`     | `--effort max` (Opus only) | `-c model_reasoning_effort=xhigh`  |

  **Note:** Claude Code's `max` is Opus 4.8 only. Codex has additional levels (`none`, `minimal`) not exposed via AI Runner.

  Combine with other flags:

  ```bash theme={null}
  ai --codex --high --effort max task.md
  ai --aws --opus --effort high task.md
  ```

  Save as default:

  ```bash theme={null}
  ai --effort high --set-default
  ```
</ParamField>

## Provider Profile

<ParamField path="--profile" type="flag">
  **Select a config profile** — Codex CLI only

  ```bash theme={null}
  ai --codex --profile azure task.md        # Azure OpenAI
  ai --codex --profile openrouter task.md   # OpenRouter
  ```

  Selects a named profile from `~/.codex/config.toml`. Use this for providers that aren't built into AI Runner's flag system (OpenRouter, Mistral, DeepSeek, xAI, Groq, etc.).

  **Setup** — configure profiles in `~/.codex/config.toml`:

  ```toml theme={null}
  [profiles.azure]
  model_provider = "azure"
  model = "gpt-5.3-codex"

  [model_providers.azure]
  name = "Azure OpenAI"
  base_url = "https://YOUR_RESOURCE.openai.azure.com/openai/v1"
  env_key = "AZURE_OPENAI_API_KEY"

  [model_providers.openrouter]
  name = "OpenRouter"
  base_url = "https://openrouter.ai/api/v1"
  env_key = "OPENROUTER_API_KEY"
  ```

  See the [Codex CLI docs](https://developers.openai.com/codex/cli) for the full config reference.
</ParamField>

## Examples

### Save Runtime as Default

```bash theme={null}
# Always use Codex with the flagship model
ai --codex --high --set-default

# Always use Claude Code on AWS
ai --cc --aws --opus --set-default

# Clear — back to auto-detect
ai --clear-default
```

### Shebang Scripts

```markdown theme={null}
#!/usr/bin/env -S ai --codex --high --effort high
Deep analysis of this codebase architecture.
```

```markdown theme={null}
#!/usr/bin/env -S ai --cc --opus --effort max
Maximum reasoning for this security audit.
```

### Combining Flags

```bash theme={null}
# Codex with Azure OpenAI profile and high effort
ai --codex --profile azure --effort high task.md

# Claude Code with AWS, Opus, and auto permissions
ai --aws --opus --effort high --auto task.md
```

## Related Pages

<CardGroup cols={2}>
  <Card title="Runtimes" icon="shuffle" href="/concepts/runtimes">
    Full cross-interpreter compatibility reference
  </Card>

  <Card title="Provider Flags" icon="cloud" href="./provider-flags">
    Switch between cloud providers
  </Card>
</CardGroup>