Skip to main content
v1.2.22 makes ADE’s Claude chats behave like Claude Code, ends a credential-refresh storm that could rate-limit usage stats, and smooths mobile chat input.

​
Desktop and CLI

  • Claude chats behave like Claude Code. ADE’s Claude Agent SDK chats now match Claude Code’s behavior end to end: no more leaked duplicate claude --resume background process, correct subagent counts (no phantom subagents or interrupt card pile-ups), and chats no longer stop at turn boundaries.
  • No more usage refresh storm. A stale credentials file could trigger a token-refresh storm that got the Claude usage endpoint rate-limited. Credential reads are now cached and failed refreshes are backed off, keeping usage stats reliable.

​
iOS

  • Composer parity. The mobile Work composer picks up the same typed-trigger behavior as desktop, keeping chat input consistent across devices.
  • Keyboard dismissal. Fixed chat keyboard dismissal so the composer gets out of the way cleanly.