SHIPPED · v1.45.0

Agent Orchestration Dashboard

Agent Orchestration Dashboard - Route your work to the right AI agent automatically — a new Orchestration Dashboard lets you decide which agent handles each kind of task (planning, implementing, testing, reviewing, browser automation, docs, research, debugging) - Your choices are woven into every agent's instructions, so when you ask one agent to "get this reviewed by another agent" without naming one, it picks the agent you assigned - Add your own task categories, pin a preferred model, and leave notes for each agent - Prefer to stay in control? A "Suggest only" mode has agents propose the routed agent and wait for your go-ahead instead of delegating on their own - See every delegation in one place — a Runs view shows each time one agent handed work to another: which agent, what category, how long it took, the output size, and whether it finished, failed, or timed out - Filter by agent, status, or time range, with per-agent error counts at a glance - Optional content capture (off by default, stored only on your machine) lets you read the exact prompt and output behind any run - Track cost and dig into logs — view token usage and cost per agent, tail the orchestration logs, and jump straight to a run from its sub-agent badge - Open it from Settings → AI → Orchestration, the Command Palette, or the clock icon on any sub-agent badge

Terminals That Survive Updates - Keep your terminals running through app updates, crashes, and restarts — a new experimental setting hosts your terminal processes in a background helper, so reopening 1DevTool reconnects to your still-running shells and AI agents with full scrollback instead of restarting them - Turn it on in Settings → Terminal → "Keep terminals alive across updates & restarts"; it takes effect on the next launch

Terminals Inside Reader Mode - Run a terminal right next to what you're reading — while reading a Markdown file in Reader Mode, dock any of the project's terminals to the side to run commands or chat with an AI agent without leaving the reading view - It's a real, live terminal — full output, scrollback, and Agent Input — not a static copy - Closing the dock hands your scrollback back cleanly so nothing is lost