AI accessibility is a practical guide that explains how to support people with disabilities, how to use AI to make content accessible, how to choose accessible AI tools, and the areas where AI still has limitations.
Each effort level of the Perplexity Agent API reduces cost while maintaining or improving research quality. Additionally, a new Ultra level, enabling the most extensive and deep research, has been added.
This guide explains the mechanisms of AI agent orchestration and compares centralized, distributed, and hierarchical architectures. It also introduces seven best practices for handling failures, approval flows, and retry strategies in operations.
Perplexity's teammate adjusts AI agents, manages ongoing projects, and oversees operations and marketing. In July 2025, it first introduced a web-compatible AI assistant, offering personal agent functions such as restaurant reservations and shopping. Since then, the model’s capabilities have improved dramatically, and it is used by millions.
The Perplexity Decisions API performs a web search before answering, providing accurate responses based on the latest facts. While the AI decision model is fast and cost‑efficient, it only knows its training data and user input, so it cannot adapt to changing information. With this update, it can now handle timely questions in real time, such as the current status of a CEO or product release information.
This guide focuses on the security of the MCP (Multi-Tool Chat) system, explaining 11 representative risks ranging from indirect prompt injection to local server compromise. Practical defensive measures for each risk are also provided, organized into three categories: LLM manipulation, data and access protection, and infrastructure vulnerabilities.
An autonomous agent is an AI system that plans goals set by the user, searches and analyzes information using tools, and presents summarized results. The article explains the processes of planning, tool usage, memory, and verification, and introduces examples of its application in research, email, tasks, and shopping.
Computer for e-commerce is a tool that supports brand growth through Shopify sales analytics, product page creation, and the development of advertising and email campaigns. It streamlines important tasks such as market research, storefront design, advertising, and email marketing, allowing business owners to focus on product development and customer service.
AI agent memory is a system where episodic, semantic, and procedural memories work together, allowing the agent to self-improve through continuous context. This article explains its structure, operating principles, and the steps to build and test an agent.
BrowseSafe++, released by the Perplexity Secure Intelligence Institute, offers an open benchmark and lightweight detector for browser agents. It detects indirect prompt injection from malicious text placed on web pages, preventing incorrect answers and unauthorized actions.