Repository navigation
user cli
This reference assumes steiner is installed. See Installation for release binaries.
Interactive mode launches the TUI:
steinerSingle-shot mode runs one request and exits:
steiner --exec "explain the auth package"Select a raw provider/model reference for a one-shot request:
steiner --model ollama/qwen2.5-coder:14b --exec "explain the auth package"The same selection can be set for the active orchestrator with STEINER_MODEL:
STEINER_MODEL=ollama/qwen2.5-coder:14b steiner --exec "explain the auth package"| Command | What it does |
|---|---|
version |
Print build version; --short prints script-friendly output |
config |
Print resolved configuration |
login codex |
Authenticate with OpenAI Codex via OAuth |
login codex status |
Show current Codex authentication status |
model inspect <alias> |
Show resolved configuration for a model alias |
models refresh |
Refresh model discovery caches |
models status |
Show model discovery cache status |
cache refresh |
Refresh model metadata and provider model caches |
oneshot [task] |
Run an autonomous oneshot task; --list, --resume <id>
|
update / upgrade
|
Self-update to the latest release |
tools |
List configured tools and approval status |
skills |
List discovered skills |
worktrees --list |
List delegation worktrees |
worktrees --prune <id> |
Remove one delegation worktree |
worktrees --prune-all |
Remove all delegation worktrees |
--help |
Print all flags and usage |
steiner --version and steiner version print build metadata. Use steiner version --short to print only the raw version string for scripts and CI.
Common flags; steiner --help lists all of them.
| Flag | What it does |
|---|---|
--config <path> |
Use a specific project config file |
--model <ref> |
Override the active orchestrator model |
--profile <name> |
Select a model profile |
--exec |
Run a single request and exit |
--unsafe |
Disable sandboxing (bubblewrap) for tool execution |
--trust-project-config |
Trust this project's config for this run without prompting |
--verbose |
Enable verbose logging |
--log-file <path> |
Write full session logs to a file |
--dev |
Select the dev channel for steiner update
|
Type / to open the command and skill list, and press Ctrl+F1 to toggle the help overlay with keybindings.
Use /compact or /compact <focus text> to compact the conversation. Use /model to change the active orchestrator model. For example, enter /model ollama/qwen2.5-coder:14b to select a raw reference. Use /profile <name> to select a profile for future role assignments. Use /mode [plan|build] to switch execution mode. See Execution modes.
A configured alias is an optional alternate for a model with persistent customization, such as --model careful when careful is defined under models.definitions with reasoning or retry settings. Raw references are the normal path when no such customization is needed.
For the full configuration and flag reference, see Configuration. For provider setup, see Provider and model setup.
Get started
Using Steiner
- How to work with steiner
- Interactive mode
- CLI reference
- Skills
- Sub-agent delegation
- Oneshot
- Context management
- Advisor
- Security audit
Tools and integrations
Safety
Usage and cost
Reference
Internals and maintenance