> ## Documentation Index
> Fetch the complete documentation index at: https://docs.prisme.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Managing Collections
> Create, organize, and delete your data collections
The **Collections** page lists all your datasets. From here you create new collections, search existing ones, and open a collection to work with its data.
## The Dashboard
The **Dashboard** gives you a quick health check before diving into a specific collection:
* **Collections** — how many collections you own or have access to
* **Total Rows** — the combined number of records across all collections
* **Queries (7d)** — how often agents and users queried your data in the last 7 days
* **Activity (7d)** — recent write operations on your data
Below the metrics, the **Recent Collections** grid shows your six most recently used collections. Click **View All** to open the full list.
## Viewing Collections
Each collection appears as a card showing:
* Its **name** and **description**
* The number of **rows** and **columns**
* The **agents linked** to it, when applicable
Use the **search bar** to filter collections by name or description as you type.
## Creating a Collection
Click **New Collection** (from the Dashboard) or **Create** (from the Collections page).
Enter a **name** (required) and a **description** (optional). A clear description helps teammates — and agents — understand what the data represents.
* **Empty (add rows later)** — creates a blank collection you can fill later, manually or through an agent
* **Upload CSV** — imports a file and detects the schema automatically (see [Importing Data](/products/ai-collection/import-data))
Click **Create**. Your collection appears in the list, ready to use.
Start from a CSV whenever you can: the schema (column names and types) is inferred from your actual data, which is faster and less error-prone than defining it by hand.
## Opening a Collection
Click any collection card to open its detail page. The header shows the name, description, and size (**X rows • Y columns**), and the content is organized in tabs:
| Tab | Purpose |
| ------------- | --------------------------------------------------------------------------------------------------- |
| **Data** | Browse, sort, and manage records — see [Working with Data](/products/ai-collection/data-management) |
| **Visualize** | Charts built from this collection — see [Visualizations](/products/ai-collection/visualizations) |
## Deleting
### Deleting a Collection
Click the collection card, or open the **⋯** menu on the card in the list.
Click **Delete** and confirm.
Deleting a collection permanently removes the collection **and all its records**. Agents linked to it lose access to the data. This cannot be undone.
### Deleting Records
To remove individual rows without deleting the collection, select them in the **Data** tab and use the **Delete** action — see [Working with Data](/products/ai-collection/data-management).
## Settings
The **Settings** page shows the technical information agents and integrations need:
* **MCP Endpoint** — the URL agents use to access your collections (`/ai-collection/mcp`, JSON-RPC 2.0)
* **Usage** — your collections count, total rows, and query volume
## Best Practices
Keep each collection focused on a single kind of record — deals, contacts, tickets. Mixed datasets are harder to query and produce confusing aggregations.
The description is visible to agents when they discover your collections. "Q3 2026 sales pipeline, one row per deal, amounts in EUR" tells an agent exactly how to use the data.
Aggregations group by exact values. "In Progress", "in progress", and "In-Progress" become three different groups — normalize your categories before importing.
Old collections still appear in agent tool listings. Deleting stale datasets keeps agent behavior predictable.
## Next Steps
Upload a CSV with automatic schema detection
Browse, sort, and manage records