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

# Environments & tenancy

> How to use environments to isolate staging vs prod and per-customer data.

Environments are Memic's tenancy boundary. Every file, every vector, every
search query is scoped to exactly one environment, and there is no way to
query across environments from a single API key.

If you take one thing away from this guide: **environments are the unit of
isolation**. Whatever you need to keep separate, put it in different
environments.

## Common patterns

### Staging vs production

The simplest use: two environments, two keys, same codebase.

```python theme={null}
import os
from memic import Memic

client = Memic(api_key=os.environ["MEMIC_API_KEY"])
```

```bash theme={null}
# dev
export MEMIC_API_KEY="mk_live_staging_..."

# prod
export MEMIC_API_KEY="mk_live_prod_..."
```

### Per-customer isolation

If you're building a multi-tenant SaaS where each customer has their own
knowledge base, create one environment per customer.

Your provisioning flow:

1. Customer signs up
2. Your backend creates a new environment in Memic (via the dashboard or
   admin API)
3. Your backend creates an API key scoped to that environment
4. You store the key in your database, associated with the customer

Every time you handle a request from that customer, you load their specific
API key and instantiate a Memic client. Customer data is never visible to
other customers because the key cannot access other environments.

See the [Per-customer isolation recipe](/recipes/per-customer-isolation) for
a full walkthrough.

### Feature branches

If you're running many feature-branch deploys on a staging server, create a
short-lived environment per branch. Delete it when the branch merges.

## What's inside an environment

Every environment has:

* Its own **file namespace** — files are not shared across environments
* Its own **vector index** — search never crosses environment boundaries
* Its own **prompts** — prompt names are scoped per environment
* Its own **folders** — folder names are scoped per environment
* Its own **connectors** — if you have connector-backed structured search

Organizations, projects, and users are **shared across environments** — they
sit above environments in the hierarchy.

## Switching environments in code

Because environments are key-scoped, switching is just swapping the key:

```python theme={null}
staging_client = Memic(api_key=STAGING_KEY)
prod_client = Memic(api_key=PROD_KEY)
```

You cannot switch environments mid-request. One client, one environment.

## Creating and managing environments

Environments are created and managed from the Memic dashboard. Navigate to
your project → Environments → Create.

## Related

<CardGroup cols={2}>
  <Card title="Core concepts" icon="diagram-project" href="/core-concepts">
    The full hierarchy explained.
  </Card>

  <Card title="Recipe: Per-customer isolation" icon="users" href="/recipes/per-customer-isolation">
    Multi-tenant SaaS pattern.
  </Card>
</CardGroup>