Perplexity’s new AI agent can perform multi-step tasks on your Android device

By Andrew Tarantola Published January 23, 2025

Perplexity announced Thursday that it is beginning to roll out an agentic AI for Android devices, called Perplexity Assistant, which will be able to independently take multi-step actions on behalf of its user.

“We are excited to launch the Perplexity Assistant to all Android users,” Perplexity CEO Aravind Srinivas wrote in a post to X on Thursday. “This marks the transition for Perplexity from an answer engine to a natively integrated assistant that can call other apps and perform basic tasks for you.”

The Assistant will be available through the Perplexity mobile app and will run atop the platforms existing “answer engine” model. As such, Assistant will have access to the internet. With it, users will be able to set reminders and future actions, much like ChatGPT‘s new Tasks feature offers. For example, the agent will be able to remind users of an upcoming event by automatically creating a calendar entry at the correct time and date.

Users can also use it to take more immediate action such as hailing a ride share or searching for a song, the company noted. The new feature can also access the user’s camera so you could, in theory, ask it to look for restaurants in your immediate area and then have it make reservations for you.

Perplexity Assistant is free to use as part of the mobile app and will initially be available in 15 languages, including English, Spanish, French, German, Japanese, Korean, and Hindi. How well it will interact with other agentic AIs on the device, such as Gemini or ChatGPT Tasks, remains to be seen.

Agents are the hot new thing in generative AI. These lightweight models are often “distilled” from larger LLMs like ChatGPT, Claude, or Gemini, but are tasked with  interpreting data and autonomously taking action rather than generating content. These actions can be straightforward, like automatically transcribing a Zoom call, or multi-step — think, having it plan an 8-course meal, shop for necessary ingredients on Instacart, then email invites to your guests.

The market is already being saturated with AI agents from the various leading companies. Anthropic kicked off the agentic race in November when it debuted its Computer Use API, which enables Claude to emulate human mouse and keyboard actions to control the local computing system.  Microsoft announced Copilot Actions the same month and began rolling the agents out to business and enterprise subscribers in January. Nvidia followed suit at CES 2025 when it revealed its new Nemotron family of LLMs, and OpenAI finally unveiled its AI agent, Operator, as a “research preview” just a few hours ago.