FAQ & troubleshooting
Real questions from real testers, plus the things to check when something misbehaves.
Frequently asked
Do I have to register a repository to use the work board?
No. The board starts from your Jira tickets and finds pull requests by ticket key across GitHub — a ticket with no branch anywhere is exactly the one it should show you. Registering a repo only adds the local branch, its tests and its captures to the row.
The board shows tickets but no pull requests (or the other way round).
They're separate sources and each says why it's missing. Tickets need the Jira site, your email and an API token on the active profile (Profiles → Jira). Pull requests need gh installed and signed in — Settings → Tools installs it. If pull requests you expect are missing, put your organisation in GitHub owners next to the query: without one the search is deliberately scoped to work you're involved in.
Does a pipeline need Docker?
Only if it declares services. Such a pipeline refuses to start when Docker is missing or its engine isn't running, and says which — a run that fails halfway for want of a database has wasted everything before it. Pipelines without services need nothing extra.
A pipeline step failed. Do I have to run the whole thing again?
No. Edit the step on the canvas and press ↻ on it: it runs again with the pipeline as it is saved now, everything downstream re-runs with it, and everything before it keeps its work and its commits.
How do I make the agent read a specific folder?
The agent works in the directory it was launched in. Pick that folder in the Launch dialog — a registered repo, Custom folder…, or a scratch folder. To hand it extra files mid-session, drag them onto the window or paste a screenshot: AIOAT types the file path into the session.
I already have Claude Code subagents (.claude/agents/*.md). Can I use them without recreating anything?
Yes — they're just files. Put them in your repo's .claude\agents\ (or your user-level %USERPROFILE%\.claude\agents\) and every Claude session AIOAT launches picks them up automatically; /agents inside the session confirms it. If they live on another machine, copy them over (scp/WinSCP) — and see the Importer to bring them into AIOAT's shared context too.
My whole setup lives on my server. Do I start from scratch on Windows?
No. Either copy your environment over (.claude/, CLAUDE.md, a git clone of the project), or don't move it at all: add the server under Servers and run the agents there — tmux keeps them alive across disconnects, and share context hands them your local workspace's shared knowledge at launch. Step-by-step: Workflows → Run your agents on your own server.
What's the “Location” dropdown in the Launch dialog? It only shows “Custom folder…”.
It lists your registered repos — empty until you add one under Code → Repositories. Registered repos unlock the git flow (worktrees, branches, diffs, review). “Custom folder…” and “No repo (scratch)” work without any of that.
Can I ask the docs instead of reading them?
Yes — hit 💬 Ask the docs in the in-app Docs pane (or palette → Help: Ask the docs…): a chat opens next to the docs, and your agent answers from the relevant documentation excerpts. Docs only — usage, not internals — in whatever language you ask in, follow-up questions welcome.
Does closing AIOAT kill my agents?
No — sessions live in the detached session host and reattach when you reopen the app. Killing a session explicitly (Kill button / removing with kill) does end its process.
Can agents see my other profile's keys?
No. A session only gets the environment of the profile it was launched under; secrets stay in Windows Credential Manager and are injected per-process, never written to disk or made global.
Troubleshooting
| Symptom | Check |
|---|---|
| Status bar shows ○ Host | The session host connects when the first terminal opens. If it stays ○ after opening one, quit AIOAT, end any aioat-session-host.exe in Task Manager, reopen. |
“claude is not installed (not found on PATH)” | Install it via Settings → Agents (or the welcome screen). If you just installed it in a terminal, restart AIOAT so it re-reads PATH. |
| Agent launches but immediately asks to log in | Log the CLI in once from any shell tab (claude → follow the prompts). Logins are per-machine, not per-AIOAT-profile. |
| Wrong-account guard fires unexpectedly | The repo's origin doesn't match the profile's allowed remote patterns — edit the patterns in the profile, or launch with the intended profile. |
| Issue fetch fails | GitHub: gh auth status must pass for the profile's account. Jira: set the site URL, email, and API token on the profile. |
| A repo doesn't show in the sidebar | The Repositories list is workspace-scoped. Edit the workspace and tick the repo — or add it while that workspace is active. |
| Merge queue stopped | By design: a failed rebase or red test run halts the queue. Open the entry's log, fix or discard, re-queue. |
| Something else | Logs live at %LOCALAPPDATA%\dk.imjia.aioat\aioat.log — include the tail when reporting a bug. |
Upgrading
Install the new MSI over the old version — settings, profiles, repos, workspaces, context and history carry over. Upgrading from 0.2.x: the app identity changed to dk.imjia.aioat; your data migrates automatically on first launch.