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

# Asynchronous Task Execution

> Handle long-running Logic workflows efficiently with async MCP support

## Overview

Logic MCP servers support asynchronous task execution, allowing AI platforms to start long-running tasks without blocking while they complete.

Instead of waiting for completion, the AI platform receives an execution ID and can check back later for results.

## How It Works

When an AI platform executes a Logic agent through MCP:

1. **Task starts immediately** - Logic begins processing
2. **Execution ID returned** - AI platform receives a unique identifier
3. **Status checking** - Use `check_execution_status` tool with the execution ID
4. **Results retrieved** - When complete, the tool returns the output

### Example Flow

```
AI Platform → Start Logic agent execution
Logic MCP → Returns execution_id: "exec_abc123"
AI Platform → check_execution_status("exec_abc123")
Logic MCP → Status: "running"
[wait a bit]
AI Platform → check_execution_status("exec_abc123")
Logic MCP → Status: "completed" + results
```

## Running Multiple Tasks

The async model enables parallel execution. AI platforms can:

* Start multiple Logic agents simultaneously
* Poll each execution ID separately
* Retrieve results as each task completes

**Example**: Process 10 images in parallel instead of sequentially

* Without async: 10 × 30 seconds = 5 minutes total
* With async: \~30 seconds total (parallelized)

## Platform Support

Different AI platforms handle async tasks automatically:

* **Claude platforms**: Automatically poll and manage multiple executions
* **ChatGPT**: Check platform documentation for async behavior
* **Custom clients**: Implement `check_execution_status` tool and polling logic

## Next Steps

* [Review agent scope](/mcp/agent-scope) - Control which agents support async execution
* [Explore integrations](/mcp/integrations/claude-desktop) - See how platforms handle async tasks