> ## Documentation Index > Fetch the complete documentation index at: https://docs.stardeck.ai/llms.txt > Use this file to discover all available pages before exploring further. # Modes & Settings > Starcat's chat modes, agent tiers, and conversation settings, and when to use each When you chat with [Starcat](/starcat), you pick a mode and an agent tier from the selector in the chat input. The mode decides how much Starcat is allowed to do; the tier decides how much reasoning power it brings. ## Modes Starcat runs in one of four modes. Agent mode is the default. | Mode | What it does | Reach for it when | | ---------- | ---------------------------------------------------------------------------------- | ------------------------------------------------------------------- | | **Agent** | Takes action: runs tools, writes data, and dispatches projects. | You want Starcat to make the change, not just talk about it. | | **Plan** | Read-only. Researches and drafts a plan for your approval before anything changes. | The task is non-trivial and you want to review the approach first. | | **Design** | Iterates on visual designs and HTML preview mockups, without touching your data. | You're mocking up a layout or exploring a look before building. | | **Chat** | Answers questions without making any changes. | You just want to ask about your business, knowledge store, or apps. | In Agent mode you can also turn on **developer agents** so Starcat can create apps and dispatch work to projects. See [Allowing developer agents](#allowing-developer-agents) below. When Starcat Developer is building across your apps in a **workspace session**, you'll also see **Ultra** at the top of the mode list — a max-power version of Agent mode. It doesn't appear in a normal Starcat chat. See [Starcat Developer Ultra](/starcat-developer-ultra). ## Agent Tiers Next to the mode, you choose an agent tier. Starcat uses the same four tiers as [Starcat Developer](/developer-agent#agent-tiers): Lite, Core, Plus, and Pro. They trade speed and cost against reasoning depth. Starcat defaults to **Plus**. * **Lite or Core** for quick questions, knowledge-store lookups, and light edits. * **Plus** (the default) for most day-to-day work, including operating your apps and planning changes. * **Pro** for complex orchestration across several projects, or anything that needs the deepest reasoning. See [Agent Tiers](/developer-agent#agent-tiers) on the Starcat Developer page for the full breakdown of each tier. Instead of fixing a tier, you can switch on **Auto** and let Starcat pick the right one for each message. When you're working in another language, the **Multi-lingual** option (shown with Core selected) uses a model that's stronger at non-English writing and visual input. See [Auto mode](/auto-mode) for both. ## Ultra **Ultra** is a max-power version of Agent mode that runs Starcat Developer with parallel write agents and end-to-end self-validation. On the organization side it appears in **workspace sessions**, where it also adds a per-run credit **budget**. It's the setting to reach for on a substantial change you want built, verified, and handed off in one go. Ultra is in alpha and rolls out gradually. For the full breakdown — including how to prompt it — see [Starcat Developer Ultra](/starcat-developer-ultra). ## Conversation Settings ### Chat visibility Every conversation has a visibility you set from the chat's settings: * **Personal** — only you can see the chat. This is the default. * **Shared with org** — anyone in your organization can see it. Visibility is per-chat, and you can change it at any time. The toggle works the same whether you're in the standalone dashboard or the Starcat sidebar next to an app you've opened. Use **Shared with org** for work your team should be able to follow or pick up. ### Allowing developer agents By default Starcat answers questions and operates your existing apps. Building new apps and dispatching work to each project's Starcat Developer is gated behind a per-conversation toggle. Turn on **developer agents** for a chat when you want Starcat to coordinate changes across your projects. Operating your apps' APIs doesn't require this toggle; it's controlled separately by each app's **Allow agent access** setting. ## Related The max-power agent mode: parallel writers, self-validation, and a spend budget What Starcat is and everything it can do Let Starcat pick the right tier for each message automatically Tips for getting the most out of Starcat