Software Engineer

Perplexity Portable Computer runs AI agents locally with zero token charges. Perplexity has launched Portable Computer, an on-device AI agent that keeps models, files, tools, and workflows on local Nvidia hardware instead of sending every task to the cloud. Local processing avoids token charges and limits data exposure, while users can approve individual handoffs to larger cloud models when the local system cannot finish a task. Local agents on Nvidia hardware. Portable Computer is initially available for Perplexity Pro and Max subscribers using Nvidia's DGX Spark desktop system, which starts at roughly $5,000. DGX Spark combines Blackwell-based graphics hardware with a 20-core CPU and 128 GB of memory, providing the capacity needed for local multi-step work such as research, coding, document analysis, and web-application development. The software uses Qwen 3.8 27B or Perplexity's DGX Spark-tuned PPLX 27B model. Perplexity also added multi-token prediction to improve processing speed and context compaction to summarize oversized prompts when real-world workloads approach the model's practical limit of about 100,000 tokens. Permission-based cloud escalation. Portable Computer begins tasks locally and can stop to request permission before sending a particular step to a more capable frontier model in the cloud. It can also connect to services such as GitHub, but external tools require user approval and are not given unrestricted access to local files. For system administrators evaluating the agent, the isolation features are as important as the local inference. A sandbox restricts model access to unrelated operating-system areas, while network controls block unauthorized connections. Built-in skills target research, data science, and coding, and a response-verification mechanism is intended to catch inaccurate results. Windows RTX support is next. The initial hardware requirement limits adoption to DGX Spark owners, but Perplexity plans to support Windows PCs equipped with Nvidia RTX graphics cards. An upcoming release is also expected to add Nvidia's 30-billion-parameter Nemotron 3.5 Lightning model, designed for capable inference on more constrained devices.

Allium brings onchain data to Perplexity Computer. Allium is partnering with Perplexity to bring Allium's onchain data directly into Perplexity via new MCP connector for Perplexity Computer, its agentic platform for multi-step research and analysis. Allium is partnering with Perplexity to bring Allium's onchain data directly into Perplexity through a new MCP connector for Perplexity Computer, its agentic platform for multi-step research and analysis. That means financial professionals can ask the questions they already care about and use Allium's data to understand what's happening onchain. Questions like: Where is perps open interest concentrated today? Which stablecoins saw the largest supply changes this week, and who's holding them? Which companies are issuing tokenized funds onchain, and how fast is that supply growing? Instead of finding the right dataset, writing a query, and interpreting the results separately, users can start with the question they actually want answered. Bringing onchain markets into the research workflow. More financial activity is moving onchain, from stablecoins and payments to tokenized assets and new market structures. For financial professionals, the challenge isn't accessing the blockchain. The data is public. The challenge is turning billions of raw transactions into information they can actually use. Allium does that by indexing, decoding, labeling, and normalizing onchain activity across 150+ chains, providing a consistent data layer for financial analysis. Allium is SOC 1 and SOC 2 compliant, with data already used by institutions including Visa and Grayscale. With the new connector, that infrastructure becomes accessible directly through Perplexity Computer: ask the financial question, and Perplexity can use Allium to retrieve the underlying onchain data and synthesize the answer. "Onchain data may be public, but making it useful still requires specialized infrastructure and expertise. By bringing Allium's normalized data into Perplexity Computer, financial professionals can ask the questions they care about in plain language and get answers grounded in the underlying onchain record. Computer and Allium are connecting trusted data with agentic research so people can move from question to insight faster." - Jeff Grimes, Product Manager for Perplexity Computer Data agents can rely on. As financial research becomes increasingly AI-assisted, getting an answer quickly is only part of the equation. You also need confidence in the data behind it. That makes the underlying data layer (its coverage, definitions, methodology, and provenance) increasingly important. Allium was built to provide that foundation: normalized onchain data with the infrastructure and methodology required for institutional workflows. It's the same principle behind its broader approach to financial data: the interface can change, but the underlying record still needs to be reliable. The integration can also return links back to the underlying Allium data, giving users a path from an AI-generated answer back to its source. Getting started. The Allium connector is available in Perplexity Computer, bringing Allium's onchain data directly into Perplexity's research experience.

Chronograph, a provider of portfolio monitoring and analytics technology for private capital investors, has partnered with Perplexity to integrate its data into Perplexity's answer engine and agent platform, Perplexity Computer. The integration allows private capital professionals to query validated portfolio data in plain language within Perplexity alongside licensed research sources. The connector enables teams to pull portfolio data from Chronograph for analytics and reporting, with all figures traceable to source documents. Chronograph is now available amongst Perplexity's Licensed Finance Data Sources. The connector is available to Chronograph clients with eligible Perplexity Pro, Max, or Enterprise subscriptions. Customers authenticate their own entitlements without requiring a separate data contract. Founded in 2016, Chronograph serves institutional limited partners and general partners, with backing from Sixth Street, Summit Partners, Carlyle AlpInvest, and Nasdaq.