Bug report: A2A-created child agent is absent from...
# bugs-forum
TLDR: I'm losing the ability to see and manage sub agents within the Traycer gui. I was originally on 1.10 desktop app and was having this issue, so I tried updating to the latest release candidate (host files + appimage) and I'm still having the same issue. If there's a recommended fix let me know. Sometimes reloading or force reloading the window fixes it, but most of the times not.
f
a
hmm I have two epic projects going right now, in one of them that revealed a chat that was previously hidden, in the other, and the one I made this bug report for it revealed nothing new, and I still only see one agent, the main orchestrator agent, but still not the sub-agents
I also can't open any new artifacts for handoff files they create for me via the gui, i have to manually ask them to save the file locally
f
Thanks for reporting. Was able to reproduce it. Happens when the main repo is moved, renamed, deleted, or re-cloned. We'll fix this shortly
f
Thanks for sharing the report.
@elegant-airline-52152 it would be great if you can ask an agent (in the same task) to collect some diagnostic info using this prompt. It writes a single
traycer-diagnostic-digest.md
to your Desktop with everything sanitized, please share it here (or privately in the ticket).
Copy code
I'm running you INSIDE the epic that has the problem: my child agents are missing from
this epic's sidebar. Your own context tells you this epic's id — record it in the
digest. Note that YOU are a new agent in this epic: record your own agent id too, and
exclude yourself from every count below.

Create ONE file, `traycer-diagnostic-digest.md`, that I can safely share publicly.
Privacy rules are strict — apply them to everything you write:

PRIVACY RULES
- The digest may contain ONLY: version numbers, file timestamps, counts, yes/no facts,
  random ids (epic/agent UUIDs), log lines matching the patterns listed below, and my
  written answers. Nothing else from my machine goes into it.
- Replace my home directory prefix with `~` in every path you include.
- Replace every user id with stable tokens (USER_A, USER_B, …) and every host id with
  HOST_A, HOST_B, … — consistent within the digest, real values never included.
- Keep epic/agent/chat UUIDs as-is (they are random ids the team needs), but never
  include their TITLES, message content, or artifact content.
- No username, email, machine name, or IP addresses anywhere.

COLLECT (adapt commands to my OS; I'm on Windows, so PowerShell where needed):
1. Versions & update timeline — `traycer --version`; the installed host version and
   the modification times of the host install directory; note anything showing the
   host updated on or after 2026-08-20. Ask me for the desktop app version if you
   can't read it.
2. Agent listing comparison — run `traycer agent list` AND `traycer agent list --all`.
   For this epic, report each run's row count and a tree of: agent id, parent id,
   ownerUserId token, archived yes/no (excluding yourself). State explicitly which
   rows appear only under --all.
3. Host log evidence — from the Traycer host log (`~/.traycer/host/host.log`, plus any
   rotated siblings):
   a) every line containing "skipping row that fails canonical validation" (sanitized);
   b) the count of lines mentioning 401 or host-credential since 2026-08-20, plus 3
      representative sanitized lines;
   c) host startup lines since 2026-08-20, to show restarts/updates.
4. Process census — how many Traycer host processes are running right now, with start times.
5. Workspace facts — for each of this epic's workspace folders: exists yes/no, is a git
   repo yes/no. Nothing else.
6. Ask me and record my answers: Is the MAIN agent visible in the sidebar? Any children
   (besides you)? What does the empty area say word-for-word, and is there any error
   banner or Retry button? Did the missing children EVER show up — and if they vanished,
   did the app or host update/restart around that time?

Write the digest with those six sections in order. Save it as
`traycer-diagnostic-digest.md` on my Desktop (or in my home folder if there is no
Desktop) — NOT inside any project or repository folder — and tell me the exact path.
Before finishing, re-scan the whole file for my username, home path, email, or any
title/content and redact anything you find. Then show me the file so I can review it
myself before uploading.
@elegant-airline-52152 I tried reproducing your issue in windows but could not. Are you by any chance using a 1.2.0 host with 1.1.x or lower app? Because if that's so, then that explains the issue. Its very hard for us to support the latest app, with this particular old host architecture as the entire chat backend has moved. In the prod release we have added safety so the app will prompt the user to upgrade the app when this particular app/host version combination is detected
e
@flat-river-95835 Thanks. I checked the proposed version-mismatch hypothesis locally. My installation is: - Desktop app: 1.1.10 - CLI: 1.1.10 - Host: 1.1.11 - Host install metadata also reports 1.1.11 - Host was installed on August 7, and no Host install files were modified on or after August 20 So this does not appear to be a 1.2.0 Host running with a 1.1.x app. I also ran
traycer agent list
and
traycer agent list --all
for the affected Epic. Both returned 17 rows, including the main agent and 16 children, with the same result under both scopes. The agents exist in the backend; the GUI/sidebar is still the layer where they are missing. The requested
traycer-diagnostic-digest.md
was not created or found on my Desktop/home folder. I also found no “skipping row that fails canonical validation” lines in host.log or host.log.1. The Host logs still contain 401/UNAUTHORIZED and host-credential failures since August 20. Could you confirm whether Desktop 1.1.10 + Host 1.1.11 is a supported production combination, and whether you still want me to collect the full sanitized digest from inside the affected Epic?
4 Views