Commands
Aura’s CLI is built with urfave/cli v3. The root aura command launches the interactive assistant. Subcommands provide non-interactive utilities.
| Command | Description |
|---|---|
aura | Interactive assistant (default) |
aura run | Non-interactive prompt execution (inline or stdin) |
aura models | List available models |
aura tools | List, inspect, or execute tools |
aura query | Embedding-based codebase search |
aura vision | Analyze an image or PDF via a vision-capable LLM |
aura transcribe | Transcribe an audio file to text |
aura speak | Convert text to speech audio |
aura tokens | Count tokens in a file or stdin |
aura mcp | List configured MCP servers and their tools |
aura show | List and inspect config entities (agents, modes, prompts, providers, hooks, features, plugins, skills, tasks) |
aura tasks | Manage and run scheduled tasks |
aura plugins | Manage plugins (add, remove, update) |
aura skills | Manage skills (add, remove, update) |
aura login | Authenticate with an OAuth provider (device code flow) |
aura cache | Manage the cache (clean) |
aura web | Start browser-based UI |
aura init | Scaffold default configuration |
Global Flags
These flags are available on all commands. They must appear before the subcommand name:
# Correct
aura --agent high run "Hello"
# Wrong — flag after subcommand is not recognized
aura run --agent high "Hello"
| Flag | Short | Default | Description |
|---|---|---|---|
--config | .aura | Configuration directories to merge (repeatable, left-to-right, last wins) | |
--version | -v | Print version and exit | |
--provider | LLM provider to use | ||
--agent | (auto) | Agent to use (resolves via default: true → first visible) | |
--model | -m | Model to use | |
--env-file | -e | secrets.env | Environment files to load (repeatable) |
--show | -s | false | Dump the full resolved YAML config and exit (no session started) |
--simple | false | Use simple readline-based TUI | |
--auto | false | Enable auto mode | |
--mode | Starting mode (ask, edit, plan) | ||
--system | System prompt name (overrides agent config; matches name: from prompt files in config/prompts/) | ||
--think | Thinking level (off, on, low, medium, high) | ||
--include-tools | Glob patterns for tools to include (e.g. Read,Glob,Rg) | ||
--exclude-tools | Glob patterns for tools to exclude (e.g. Bash,Patch) | ||
--include-mcps | Glob patterns for MCP servers to connect (e.g. context7,git*) | ||
--exclude-mcps | Glob patterns for MCP servers to skip (e.g. portainer) | ||
--max-steps | 0 | Maximum tool-use iterations (overrides config) | |
--token-budget | 0 | Cumulative token limit (overrides config) | |
--workdir | -w | Working directory for tool execution and path resolution | |
--resume | Resume a saved session by ID prefix | ||
--continue | -c | false | Resume the most recent session |
--output | -o | Mirror all output to a file | |
--without-plugins | false | Disable Go plugins | |
--unsafe-plugins | false | Allow plugins to use os/exec and other restricted imports | |
--debug | false | Enable debug logging | |
--print | false | Print a one-line config summary and exit | |
--print-env | false | Print resolved settings as AURA_* environment variables and exit | |
--experiments | false | Enable experimental features | |
--home | ~/.aura | Global config home (~/.aura); set to empty to disable | |
--providers | Limit to these providers (filters model listings and discards non-matching agents) | ||
--set | Set template variables (KEY=value, repeatable) | ||
--override | -O | Override config values (dot-notation, repeatable). Example: -O features.tools.max_steps=10 | |
--dry | Dry-run mode: render prints resolved agent, prompt, and input then exits; noop runs the full pipeline with a no-op provider that returns [noop] | ||
--no-cache | false | Bypass cached data and re-fetch from source |
Note: All root flags can be set via environment variables with the
AURA_prefix (e.g.AURA_AGENT,AURA_MODEL,AURA_DEBUG). Useaura --print-envto see all resolved settings as environment variables.