Summary
Theo of t3.gg gives a full overview of T3 Code, the open-source desktop app for working with AI coding agents. He explains why he and developer Julius started the project after growing frustrated with OpenAI’s closed-source Codex app and Anthropic’s Claude Code desktop app, and he admits that for months T3 Code was a worse alternative to both.
The video traces how that changed around July, as the team focused on things the major labs were not prioritizing. These include a reworked sidebar, remote control of sessions, and the ability to bring subscriptions from Codex, Claude, Grok, Cursor and others into one place. Theo says the app is approaching 550,000 users and 140,000 weekly active users.
The walkthrough covers how he works day to day: snoozing and archiving threads, scheduled tasks that rerun on a timetable, steering versus queuing follow-up messages, and an optional working section that hides threads while agents run. He also notes that models now understand context compaction and will encode instructions to survive it. A sponsor segment on the Apify MCP server, which gives agents access to data sources they cannot normally reach, appears early in the video.
It is useful viewing for developers comparing agentic coding interfaces and thinking about how to manage many parallel agent sessions.
đź“ş Source: Theo – t3․gg · Published October 29, 2026
🏷️ Format: Showcase
