# User Guide Install: [QUICK_START.md](QUICK_START.md). Config keys: [SETTINGS.md](SETTINGS.md). ## Run - Installed: `grinta` - Source: `uv run python -m backend.cli.entry` - Interactive TTY → Textual UI. Piped stdin → non-interactive (one line = one turn; use TUI for multi-turn). ## Modes (`/mode`) | Mode | Does | | --- | --- | | **Chat** | Read-only Q&A | | **Plan** | Read-only investigation | | **Agent** | Full edit + shell loop (default) | ## Autonomy (`/autonomy`) Only matters in **Agent** mode. Controls confirmation prompts, not execution policy. | Level | Behavior | | --- | --- | | conservative | Confirm shell, edits, terminal, browser, MCP, and delegation | | balanced | Confirm high-risk (default) | | full | No prompts; CRITICAL blocks still apply | Execution hardening: `security.execution_profile` in [SETTINGS.md](SETTINGS.md). ## Slash commands | Command | Purpose | | --- | --- | | `/help` | Commands list | | `/settings` | Model, API key, MCP | | `/sessions` `/resume` | Past sessions | | `/model` `/mode` `/autonomy` | HUD controls | | `/health` | Fast env check (LLM, git, execution profile); run `grinta doctor` outside TUI for full schema checks | | `/diff` `/checkpoint` `/compact` | Workspace tools | ## Safety Runs on your machine. See [SECURITY_CHECKLIST.md](SECURITY_CHECKLIST.md) for untrusted code.