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

# Tool Output Offloading

> Prevent context window bloat from large tool responses

## What is Tool Output Offloading?

When MCP tools return large amounts of data (like database query results, file contents, or API responses), they can quickly fill up the AI's context window. This leaves less room for your actual conversation and can degrade response quality.

Tool Output Offloading solves this by storing large outputs (>10KB) separately and letting the AI fetch them only when needed.

## Why Enable This?

<CardGroup cols={2}>
  <Card title="Preserve Context" icon="brain">
    Keep your context window available for what matters - your conversation, not raw data dumps.
  </Card>

  <Card title="Avoid Client Issues" icon="shield">
    Some MCP clients crash or behave unpredictably when their context window is flooded with large outputs.
  </Card>
</CardGroup>

**Enable this if your servers have tools that return:**

* Large database query results
* File contents or logs
* Verbose API responses
* Any output that regularly exceeds 10KB

## Supported Clients

| Client  | Status    |
| ------- | --------- |
| ChatGPT | Supported |

<Note>
  Many MCP clients already handle large outputs natively with their own offloading. We only enable this feature for clients that don't do it themselves. When an unsupported client connects, the feature is automatically bypassed and outputs are returned normally.
</Note>

## How to Enable

1. Go to **MCPs** and select your server
2. Open server settings
3. Enable **Tool Output Offloading**
4. Save changes