Perplexity released the feature on September 29, 2026. Runs can start on a schedule or on Slack, Gmail, Outlook, Linear, and GitHub events, with conditional filters. They can draw on Memory, Skills and connectors, and they can pause for review before consequential actions. The new product replaces Scheduled Tasks; existing tasks migrate on open, available to all Computer users. On Product Hunt it drew 187 upvotes and 2 comments. The earlier Scheduled Tasks product already ran jobs autonomously in the cloud, but each run began from scratch. Now each run can see what earlier runs produced. Agents can also watch for missed deadlines or overdue replies and flag them. That moves Perplexity closer to trigger-based workflow tools like Zapier and n8n, and to the agent products from the large model labs. For founders, persistent agents with memory across runs are becoming a standard feature of horizontal assistants. Startups selling agents for monitoring, reporting or follow-ups will need deeper vertical data, better approval controls or different pricing to stand out. Charging per run, with no cost while an agent waits, is a pricing model worth comparing against.