# User-facing CLI
You cannot press these yourself — tell the user which command to run.
| `/hosts` | Hosts (local in-process + remotes), tools, cmd, live/idle/error |
| `/session` | Current session metadata (title, links, scratchpad, path) |
| `/sessions` | Recent sessions (titles + link counts) |
| `/resume [id]` | Load conversation memory |
| `/new` | Fresh session (saves current) |
| `/title [text]` | Show or set session title |
| `/effort [level]` | Show or set reasoning effort (`low\|medium\|high\|max`) |
| `/help` | Full help |
| Alt-Enter / Shift-Enter / Ctrl-J | Multiline input |
| Enter | Submit |
| Ctrl-C | Cancel line at prompt; cancel in-flight turn while running |
| Ctrl-L | Clear scrollback and reprint the conversation (empty prompt only) |
| Ctrl-D / `/exit` | Save and quit |
Startup banner prints model, session, config path, hosts, and default host.
### Models & env (quick)
- Default model: `grok-4.5-build` (`--model` to change).
- Claude models need `ANTHROPIC_AUTH_TOKEN` or `ANTHROPIC_API_KEY` (optional
`ANTHROPIC_BASE_URL`).
- Grok / OpenAI Responses need `XAI_API_KEY` or `OPENAI_API_KEY` (optional
`XAI_API_BASE_URL` / `OPENAI_BASE_URL`; default base `https://api.x.ai/v1`).
- `.env` in cwd is loaded at startup. Full tables: `myco --help overview`.
Thinking/reasoning is always requested (default effort=`high`). The UI shows a `Thinking: …`
summary inside a unified RESPONSE section; it is stored in session history for resume/Ctrl-L
but stripped from provider requests.