> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentset.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Data Segregation

> Keep your users' data separate in multi-tenant applications

When building apps where multiple users or customers share the same infrastructure, you need to keep their data separate. Agentset supports two approaches:

* **Metadata filtering** — Tag documents with ownership, filter at query time.
* **Tenant isolation** — Full database-level separation between customers.

## Metadata filtering

All data lives in the same tenant, you tag each document with ownership metadata. When searching, you filter to only return that user's data.

This approach is easier to manage. You enforce access control in your backend by validating the user's access permissions.

<CodeGroup>
  ```typescript TypeScript theme={null}
  await ns.search("quarterly report", {
    filter: {
      $or: [
        { ownerId: user.id },
        { teamId: user.teamId },
        { visibility: "public" },
      ],
    },
  });
  ```

  ```python Python theme={null}
  client.search.execute(
      query="quarterly report",
      filter={
          "$or": [
              {"ownerId": user.id},
              {"teamId": user.team_id},
              {"visibility": "public"},
          ],
      },
  )
  ```
</CodeGroup>

See [Filtering](/search-and-retrieval/filtering) for more options.

## Tenant isolation

Tenant isolation provides full database-level isolation. Namespaces within your organization are standalone tenants.

Within a single namespace, you can further manage tenants using the `x-tenant-id` header.

### Using the tenant ID header

Pass the `x-tenant-id` header on every request to scope it to a specific tenant. You must include this header when both ingesting and searching. Queries without a tenant ID uses the namespace's default tenant ID.

<CodeGroup>
  ```typescript TypeScript theme={null}
  await ns.ingestion.create(
    {
      payload: { type: "TEXT", text: "Document content..." },
    },
    {
      tenantId: "customer_123",
    },
  );
  ```

  ```python Python theme={null}
  client = Agentset(
      namespace_id="YOUR_NAMESPACE_ID",
      token=os.environ["AGENTSET_API_KEY"],
      tenant_id="customer_123",
  )

  client.ingest_jobs.create(
      payload={"type": "TEXT", "text": "Document content..."},
  )
  ```
</CodeGroup>

Use the same header when searching to retrieve only that tenant's documents.

<CodeGroup>
  ```typescript TypeScript theme={null}
  await ns.search(
    "quarterly report",
    {
      topK: 10,
    },
    {
      tenantId: "customer_123",
    },
  );
  ```

  ```python Python theme={null}
  client = Agentset(
      namespace_id="YOUR_NAMESPACE_ID",
      token=os.environ["AGENTSET_API_KEY"],
      tenant_id="customer_123",
  )
  client.search.execute(
      query="quarterly report",
      top_k=10,
  )
  ```
</CodeGroup>

Keep in mind that queries are scoped to a single tenant. If tenants need to share data, consider metadata filtering instead or store shared content separately.

## Next steps

* [API Reference](/api-reference/endpoint/search) — Search endpoint parameters and options
* [Document Metadata](/data-ingestion/document-metadata) — Add metadata during ingestion
* [Filtering](/search-and-retrieval/filtering) — Filter search results by metadata