[CLSTR]
< Back to situation

[REVISION HISTORY]

Perplexity AI agent development and deployment

Updated 4 times since CLSTR started tracking revisions of this situation.

What changed

2026-09-25 16:46 UTC → 2026-10-21 14:12 UTC · added removed

Perplexity has expanded its AI assistant capabilities from operating system integration to local, on-device execution. Initially, the company rolled out a Personal Computer AI assistant for Windows, expanding from its existing macOS support. This version functions as a digital worker within the Windows operating system, capable of reading and editing local files, interacting with Microsoft 365 applications, and integrating with over 400 third-party connectors such as Salesforce and HubSpot. Subsequently, In partnership with Nvidia, Perplexity launched ‘Portable Computer’ in partnership with Nvidia. This iteration focuses on local Computer’, a local-first AI agent workloads, designed platform. Initially limited to run on hardware like Linux and Nvidia DGX Spark and Linux machines systems, the platform has since expanded to Windows PCs equipped with Nvidia GeForce RTX GPUs. By moving tasks from the cloud to local devices, the system aims to reduce cloud computing costs and enhance privacy. While the agent manages its own orchestration and local search index, it retains the ability to request permission to access cloud-based frontier models for advanced reasoning or real-time web navigation. RTX PRO GPUs, requiring at least 24GB of VRAM. The Portable Computer platform system runs the entire agent stack—including the orchestrator, planner, tool router, and local search index—on-device. This allows for complex tasks like document analysis index—on-device to enhance privacy and data organization without consuming reduce cloud token credits. Initial model support includes costs. It utilizes models such as Alibaba’s Qwen 3.8 27B and Perplexity’s own PPLX 27B, with support for Nvidia Nemotron 3.5 Lightning expected to follow. Expanding its hardware compatibility, expected. The platform features connectors for various productivity tools, including Microsoft Outlook, OneDrive, Google Drive, Slack, and GitHub. Perplexity has launched also introduced a Windows version of Portable Computer. This version is designed fully local mode for compatible Nvidia GeForce RTX its Mac AI agent. This update allows model execution, task planning, and RTX PRO workstations, requiring a GPU file processing to occur entirely on-device, enabling workflows like managing Notes or drafting documents in Excel without cloud transmission. This mode utilizes Qwen 3.6 or the fine-tuned PPLX Pro model within isolated sandboxes. Due to high computational demands, Perplexity recommends Apple Silicon Macs with at least 24GB 36GB of VRAM. The software is optimized for NVIDIA RTX graphics cards and AMD Ryzen AI Max processors, marking a shift toward using the PC as unified memory, while offering a direct computing layer ‘Hybrid Compute’ mode for AI agents. The system features a built-in browser and a proprietary ‘SPACE sandbox’ to perform multi-step tasks while keeping sensitive information on the device. users with less memory.

Versions

  1. 2026-10-21 14:12 UTC Perplexity AI agent development and deployment
  2. 2026-09-25 16:46 UTC Perplexity AI agent development and deployment
  3. 2026-09-14 19:33 UTC Perplexity AI agent development and deployment
  4. 2026-08-26 05:25 UTC Perplexity AI agent development and deployment
  5. 2026-08-26 00:58 UTC Perplexity AI agent development and deployment

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.