Give it tools - Perplexity

2026-02-04 · Source: perplexity.ai via Google News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

Perplexity models can extend their capabilities beyond pre-trained knowledge by integrating external tools. This functionality is enabled by listing tools in the "tools" array, allowing the model to decide when to invoke them based on the prompt. Two primary types of tools are supported: built-in tools and function tools. Built-in tools, hosted by Perplexity, include "web_search", "fetch_url", "people_search", "finance_search", and a "sandbox" for code execution, each activated by its "type". Function tools allow agents to call user-controlled code, such as internal APIs or databases. These are defined with a "name", "description", and JSON Schema "parameters". When a function tool is called, the model pauses, providing arguments for the user to execute the function and return the result, which is then replayed to the model. Tool invocation can be controlled using "tool_choice" and "parallel_tool_calls" parameters. Tool outputs, like "search_results" or "function_call" items, are returned in the response's "output" array.

Key takeaway

For AI Engineers building agents that require real-time data or custom system interactions, you should integrate Perplexity's built-in or function tools. This allows your models to perform actions like live web searches or query internal APIs, significantly expanding their operational scope. Define function tool schemas carefully and manage external execution and result re-injection to ensure seamless agent functionality.

Key insights

Perplexity models integrate built-in and custom function tools to extend capabilities beyond internal knowledge.

Principles

• Models decide tool use based on prompt.
• Function tools require user execution.
• "call_id" correlates function calls and outputs.

Method

To integrate tools, add them to the "tools" array. For function tools, define schema, execute calls externally, and replay results with "function_call_output" using the "call_id".

In practice

• Use "web_search" for live web data.
• Integrate internal APIs via function tools.
• Run code in an isolated "sandbox".

Topics

• Perplexity API
• Tool Use
• Function Calling
• Web Search
• Code Sandbox
• Agent Development
• API Integration

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Give it vision - Perplexity — Perplexity's Agent API unifies access to diverse LLM providers and specialized tools through a single interface.
• [Day 5/100] Tool Use and Function Calling, Explained from Scratch — Function calling is trained, constrained text generation parsed into structured data, not direct function execution.
• Combine Built-in Tools and Function Calling in the Gemini Interactions API — Gemini API now combines built-in and custom tools, automating orchestration and context circulation for complex agent workflows.
• Perplexity Max Debuts Multi-AI Agent Tool — Perplexity's new agent tool unifies 19 AI models for complex workflows, targeting professionals with a multimodal strategy.
• Perplexity Just Built an AI That Does Everything — Perplexity Computer offers a fully hosted, easy-to-use AI agent with extensive integrations, simplifying complex automation.
• Prompt the agent - Perplexity — Distinguish agent "instructions" (persistent rules) from "input" (turn-specific task) for precise control and grounded retrieval.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by perplexity.ai via Google News.