A Treasure AI Studio session is shaped by four building blocks: work folders carry your standing context, model tiers trade speed against reasoning depth, credits meter what each interaction costs, and the context window governs how much of a conversation the AI can hold at once. Understanding these four concepts is what lets you guide a session toward faster, cheaper, and more predictable results.
The AI selects its own tools and manages its own context within a session. Understanding these concepts lets you guide its decisions — it does not replace them.
- Signed in to Treasure AI Studio (Getting Started)
- AI Credits provisioned for your account
A work folder groups related chats and their files, and gives the AI a set of standing instructions that apply to every conversation started inside it. Instead of re-explaining your conventions in each new chat — which database to use, how to format output — you write them once in the work folder, and the AI follows them automatically.
A work folder has two fields:
| Field | What It Does |
|---|---|
| Name | Identifies the work folder in the sidebar and the work folder switcher. |
| Instruction | Standing guidance the AI follows in every chat in this work folder — coding conventions, a preferred database, output formats. Saved as a CLAUDE.md file inside the work folder and applied as the agent's instructions. |
Example: A "CDP Ops" work folder might have the instruction: "Always validate YAML before pushing. Use the marketing database by default. Format SQL with CTEs."
Work folder instructions apply only within their work folder. For guidance that should apply to every chat across your organization, an account administrator can set Custom Instructions, which are appended to the system prompt of every conversation.
Beyond these fields, each work folder keeps its own shared file space. Every chat inside a work folder reads and writes the same files, and you can browse and organize them — create folders, rename, move, and delete — from the sidebar and the work folder page. See Organizing Files in a Work Folder.
Work folders appear in the Work Folders section of the sidebar, with a count in parentheses (for example, "Work Folders (3)"). Click the folder-plus icon next to the header to open the New work folder dialog, which asks for a Name and an optional Instruction.
On Mobile, work folders are labeled Projects. They appear in a collapsible Projects section on the home screen, each row showing its chat count. Tap a project to view its chats.
Work folders are stored server-side and sync across all your platforms. A work folder created on Web is immediately available on Desktop and Mobile.
Treasure AI Studio offers three model tiers. Each maps to a specific Claude model and trades speed and credit cost against reasoning depth. The Balanced tier is the default for new conversations.
| Tier | Model | Description (shown in the UI) | When to Use |
|---|---|---|---|
| Pro | Claude Opus | "Complex reasoning and analysis" | Multi-step data analysis, building segment rules, writing workflows |
| Balanced | Claude Sonnet | "Fast and capable" | General-purpose work — the default for new conversations |
| Fast | Claude Haiku | "Quick tasks and simple questions" | Quick lookups, simple questions, conserving credits |
You can switch tiers at any point in a conversation — start with Fast for quick exploration, then switch to Pro for the complex part. The new tier applies to your next message; your conversation history is preserved. To set which tier new chats start with, use the Default Model option in Settings.
In the Korea (ap02) region, only the Balanced tier is available — Pro and Fast are not. All three tiers are available in the US, Tokyo, and EU01 regions.
Click the model selector in the chat input area to switch tiers. The current tier (for example, "Balanced") is shown on the selector.
Tap the model label in the chat input area to switch between available tiers. When only one tier is available for your region, the model is fixed.
Use Fast for simple lookups like "list my databases" or "how many rows in this table?" — it answers in seconds and uses far fewer credits. Reserve Pro for multi-step analysis, complex segment logic, and workflow generation.
Every AI interaction consumes credits. The amount depends on which model tier you use and how much the AI processes — longer conversations and tool-heavy sessions cost more. Higher tiers cost more per interaction.
| Model Tier | Relative Cost |
|---|---|
| Pro | Highest |
| Balanced | Medium |
| Fast | Lowest |
Administrators can monitor consumption and cap per-user daily spend. See Credit Usage for the usage dashboard and Credit Policy for daily limits. To check or increase your account's credit allocation, contact your Customer Success Manager.
The context window is how much of a conversation the AI can hold at once. Within a single conversation, the AI remembers everything discussed — questions asked, data explored, decisions made — so you can build on earlier messages naturally: "Now filter that to just the US region" or "Add those users to the segment we defined earlier."
As a conversation grows, it eventually approaches the model's context limit. Studio handles this automatically with auto-compact: it summarizes older messages to free space while preserving recent context and key facts, and the conversation continues without interruption. There is nothing you need to do.
In-session memory does not carry across separate conversations. To carry context across chats, use a work folder rather than pushing a single conversation to its limit. For the full picture of how compaction and memory work, see In-Session Memory.
| Issue | Solution |
|---|---|
| Cannot create a new work folder | Verify you have an active Treasure AI account with AI Studio access. On Mobile, work folders (Projects) are created on Web or Desktop |
| Model selector not visible | The selector appears in the chat input area. In regions with a single available tier (Korea), the tier is fixed and no selector is shown |
- Agent Orchestrator — How Studio plans, selects tools, and executes your requests
- Getting Started — Set up Studio on your platform
- Supported Skills Catalog — Browse available AI skills
- Security & Permissions — Configure permissions and security settings