Utilities provide auxiliary capabilities, including capability discovery, generic API invocation, configuration management, shell completion, Model Context Protocol (MCP) services, skill management, and self-diagnostics. This topic explains how to use utilities commands.

Overview

Command	Description	Example link
capabilities	Displays a list of the capabilities supported by ve-tos-cli. You can filter the list by hierarchy, group, or keyword.	capabilities
config	Initializes or manages tool configurations with multi-profile support.	config
api	Sends requests directly to TOS server‑side APIs. This command is suitable for debugging or executing operations that are not yet encapsulated by the tool.	api (raw API passthrough)
completion	Generates scripts that complete commands in the current shell.	completion
doctor	Diagnoses issues with the runtime environment, such as tool configuration, credentials, and network connectivity.	doctor
serve (MCP)	Starts the MCP service, with support for both stdio and SSE transport modes.	serve (MCP)
skill	Lists or exports skill definition files for AI agent integration.	skill

Utilities commands

capabilities

Use the capabilities command to view a list of the capabilities supported by ve-tos-cli. You can filter the list by hierarchy, group, or keyword.

Use the --view option to specify the desired output view, choosing from groups (default), text, compact, or full.

Use --group, --search, or --layer to filter the results.

Code example:

config

Use the config command to initialize or manage tool configurations with multi-profile support.

The default configuration file path is ~/.tos/config.toml. You can override this path using the --config-path flag or the TOS_CONFIG_PATH environment variable. TOS-specific configurations are written to the profile.ve-tos file.

The following fields are adjustable:

checkpoint_dir, batch_report_dir, batch_report_format, progress_enabled, max_retry_count, requesttimeout, connecttimeout, and maxconnections.

Code example:

api (raw API passthrough)

Use the api command to send requests directly to TOS server-side APIs. This method is ideal for debugging or executing operations that are not yet encapsulated by the tool.

The --request parameter supports inline JSON and file references using the file://path syntax. Configurable fields include:

method, endpoint_rule (or endpoint_kind), bucket, key, path, query, headers, and body.

The --force flag is required to execute non-idempotent operations.

Code example:

completion

Use the completion command to generate scripts that complete commands in the current shell.

Code example:

doctor

Use the doctor command to diagnose issues with the runtime environment, such as tool configuration, credentials, and network connectivity.

Supported check items include auth, config, registry, permissions, region, network, version, mcp, principles, and completion.

Use the --live-network and --network-timeout-ms flags to control the behavior of live network probing.

Code example:

serve (MCP)

Use the serve command to start the MCP service, with support for both stdio and SSE transport modes.

Code example:

skill

Use the skill command to list or export skill definition files for AI agent integration.

The results are written to the dir/SKILL.md and dir/{domain}/{skill_name}/SKILL.md files.

The tool will refuse to overwrite an existing target path.

Code example:

Torch Object Storage