Perplexity Sandbox API Agent

Perplexity's Sandbox API enables AI agents to execute code in isolated cloud environments, supporting Python and JavaScript with pre-installed packages for data analysis, visualization, and web scraping workflows.

Overview

The Perplexity Sandbox API provides isolated code execution environments for AI agents. It allows agents to write, execute, and iterate on code in secure cloud sandboxes as part of agentic workflows.

Key Features

• Isolated execution environments per request
• Python and JavaScript runtime support
• Pre-installed data science packages (pandas, numpy, matplotlib, etc.)
• File input/output capabilities
• Timeout controls and resource limits
• Integration with Agent API tool calling

How It Works

1. Agent receives a query requiring computation
2. Agent generates code via the LLM
3. Code is sent to Sandbox API for execution
4. Results (text output, generated files, charts) returned to agent
5. Agent incorporates results into final response

Use Cases

• Data analysis and visualization
• Mathematical computations
• Web scraping and data extraction
• File format conversion
• Chart and graph generation
• Statistical analysis

Integration

Used as a tool within the Agent API:

{
  "tools": [
    {"type": "web_search"},
    {"type": "code_execution"}
  ]
}

Part of the Perplexity API Platform alongside Agent API and Embeddings API.