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

# Command Help

> Inspect LoomLoom commands, flags, and the current command tree.

Use the CLI's built-in help as the source of truth for the version installed on your machine.

```bash theme={null}
loomloom --help
loomloom <command> --help
loomloom <command> <subcommand> --help
loomloom --version
```

For example:

```bash theme={null}
loomloom run --help
loomloom run precheck --help
loomloom template-spec docs --help
```

## Global flags

| Flag                    | Default           | Purpose                                                                            |
| ----------------------- | ----------------- | ---------------------------------------------------------------------------------- |
| `-s, --server <url>`    | Configured server | Override the LoomLoom base URL for this invocation                                 |
| `-t, --token <token>`   | Configured token  | Override the bearer token; prefer `LOOMLOOM_TOKEN` to avoid shell history exposure |
| `--timeout <duration>`  | `30s`             | Set the HTTP request timeout                                                       |
| `-o, --output <format>` | `text`            | Select `text` or `json`                                                            |
| `-v, --verbose`         | Off               | Write request diagnostics to stderr                                                |
| `-h, --help`            | —                 | Show command help                                                                  |
| `--version`             | —                 | Show the installed version                                                         |

Global flags can be used with subcommands:

```bash theme={null}
loomloom run list --output json --timeout 45s
```

## Staged execution commands

Programmatic official-template execution is intentionally separated into three commands:

```bash theme={null}
loomloom run validate <template-id> --file rows.jsonl
loomloom run precheck <template-id> --file rows.jsonl
loomloom run execute <template-id> --file rows.jsonl --client-request-id <id>
```

`run execute` does not replace the confirmation step in your application or agent. Run it only after the same input has been validated, prechecked, shown to the user with its estimate, and explicitly confirmed.

<Note>
  The old combined `run submit` command is hidden in this release. New integrations should use `run validate`, `run precheck`, and `run execute`.
</Note>

For machine-to-machine chaining, use [JSON output](/developer-reference/loomloom-cli/output-formats) and preserve returned IDs exactly.