Computer-Use 2.0: Multi-Agent Parallel Desktop Control
AI Engineer · youtube · 2026-07-16
This talk introduces Computer-Use 2.0: enabling multiple agents to operate in parallel on the same desktop without hijacking the user's mouse and keyboard focus.
The core idea is to sink the automation layer from "hardware mouse/keyboard events" down to lower-level system accessibility APIs: UI Automation for Windows, AT-SPI for Linux, and AX for macOS. This allows clicks and inputs to be sent directly to background windows without bringing the target app to the foreground, meaning multiple agents can work simultaneously while the user continues typing normally.
The talk will also cover:
- Overall architecture and API design
- Platform-specific quirks across Windows / macOS / Linux
- Remaining limitations: Wayland, Chromium DOM surface, native canvas apps, fallback input paths
- Live demo: three agents operating on one desktop simultaneously without interrupting user input
More from coding & agent
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- LangSmith adds tracing for Pipecat, LiveKit, OpenAI Realtime, and Gemini Live — LangChain · 2026-07-22
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- Annotated transcript of a Claude Code team interview is now available — trq212 · 2026-07-22
- Claude Code skill uses 10 Markdown rules to make outputs ADHD-friendly — alex_verem · 2026-07-22