Building Intelligent Research Agents with Manus – Ivan Leo, Manus AI (now Meta Superintelligence)

Ivan Leo, co-founder at Manus AI (subsequently joining Meta Superintelligence), leads a hands-on workshop introducing the newly launched Manus API and walking through four practical applications built with it. The demos include a personalized French language-learning app integrating 11 Labs text-to-speech, a Rick and Morty character database visualizer generated from structured JSON data, an automated analysis of Warren Buffett’s investor letter parsed directly from PDF, and a Slack bot that reproduces Manus functionality within workplace messaging.

The technical content covers core API patterns developers will need in production: file uploads, public URL ingestion, base64-encoded image processing, task status polling, and — the highlighted best practice — webhook-based task completion notifications as a scalable alternative to holding open polling loops. Leo also details the Manus 1.5 architecture improvements (speed gains, quality improvements, full model rearchitecting) and announces a new Microsoft 365 integration enabling Manus to edit Word documents, Excel templates, and PowerPoint presentations directly.

The session is aimed at developers building agentic workflows programmatically, with an emphasis on Manus’s philosophy of meeting users across every surface: web app, iOS, Slack, email, browser operator, and enterprise productivity suites. The workshop includes live coding in a Jupyter notebook environment, making failure modes and workarounds visible in real time.

📺 Source: AI Engineer · Published December 30, 2025
🏷️ Format: Hands On Build