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

# Server mode

> Share memory across agents and enable multi-tenant isolation

By default, plyra-memory runs entirely locally. Server mode connects the library to a running [plyra-memory-server](/memory-server/overview) instance, enabling:

* **Multi-agent memory sharing** — multiple agents share a memory space
* **Multi-tenant isolation** — each API key has its own isolated workspace
* **Cross-process recall** — memory survives process restarts and redeploys
* **Centralised management** — one server, multiple teams or customers

## Connect

Set two environment variables:

```bash theme={null}
export PLYRA_SERVER_URL=https://plyra-memory-server.politedesert-a99b9eaf.centralindia.azurecontainerapps.io
export PLYRA_API_KEY=plm_live_your_key_here
```

No code changes. The library detects `PLYRA_SERVER_URL` and routes all `remember()`, `recall()`, and `context_for()` calls to the server.

```python theme={null}
# Exactly the same code — works in local and server mode
async with Memory(agent_id="support-agent") as mem:
    await mem.remember("User opened ticket #4821")
    ctx = await mem.context_for("recent user issues")
```

## When to use server mode

| Situation                                      | Mode            |
| ---------------------------------------------- | --------------- |
| Single agent, development                      | Local           |
| Single agent, production                       | Local or server |
| Multiple agents sharing memory                 | Server required |
| Multiple teams or customers                    | Server required |
| Need cross-process recall without file sharing | Server          |

## Get an API key

<Card title="Generate an API key" href="/memory-server/api-keys" icon="key" color="#818cf8">
  Create a live API key in seconds. Shown once — store it securely.
</Card>

## Test the connection

```python theme={null}
import asyncio
from plyra_memory import Memory

async def test():
    async with Memory(agent_id="test") as mem:
        await mem.remember("connection test")
        ctx = await mem.context_for("connection test")
        print("Connected:", ctx.content)

asyncio.run(test())
```

## Switching between modes

```bash theme={null}
# Local mode (default — no env vars needed)
unset PLYRA_SERVER_URL

# Server mode
export PLYRA_SERVER_URL=https://...
export PLYRA_API_KEY=plm_live_...
```

Your code is identical in both modes.