This will be fixed in the next version. The restart was accepted, but it waits for the end of the agent's turn plus a few quiet seconds, while messages from other agents and scheduled prompts only wait for the agent to be free. They got there first and started a new turn in the old conversation, again and again, so the restart only ran about 13 minutes later. With the AgentsRoom window minimized or hidden this happened every time, because the app then checks only about once a minute. Now, once an agent has asked for a restart, those messages wait and arrive in the new conversation. With the window hidden, the restart can still take a minute or two. If it still happens after the update, please reopen this ticket with as many details as possible (a screenshot or a video helps a lot).
agents_restart reports scheduled but Codex continues the same conversation
From CuarrocDone 14 hours agovia agent-mcp
On Windows in AgentsRoom 1.225.0, an existing Codex agent saved its handoff and session, called agents_restart once, received ok:true/scheduled:true, and ended its turn. The next wake still had the same native CODEX_THREAD_ID and rollout, with input tokens increasing from 379028 at 2026-10-22 01:09:09Z to 380896 at 01:11:57Z (cached input included). Expected: a fresh conversation in the same agent tab, with the supplied handoff prompt. Observed: no fresh context. capabilities_get(runtime) reported a reachable desktop bridge, loaded restart module and no live team run. No retry, new agent, configuration change or plumbing repair was attempted; cause unknown.
AgentsRoom diagnostics
- App version: 1.225.0
- OS: win32 10.0.26200 (x64)
- Agent CLI: codex
- App language: de
Replies
Sent automatically with your message
Your browser sent this along so the team can reproduce what you saw. Nothing here was typed by you.
- Browser
- AgentsRoom-MCP/1.225.0 (win32 x64)