You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Thread details: see and stop the commands an agent left running (dev servers, watchers)
#34598
When an agent starts something in the background, such as a dev server, a watcher or a long test run, it's hard to keep track of afterwards:
The only sign is the background work line above the composer. Thread details, where I look for everything else about a thread, doesn't list it.
The line's Details dialog shows the command but not its output, so I can't tell whether my server started, crashed or is waiting on something.
Stop ends all of the agent's background work at once. With a backend and a frontend running, I can't stop just one.
When the command writes its output to a log file (> dev.log 2>&1, which agents do a lot), there's nothing to see anywhere in T3 Code.
On Windows, Claude commands don't stream any live output, because Claude's task output files are looked for in the POSIX temp location.
Example: I asked Claude to start the backend (.NET API) and frontend (Vite) of another project. It started both in the background, and the only way to see whether they came up was to ask the agent or find the log files myself.
Proposal
A Running section in thread details, on by default and toggled from the existing "Show in thread details" menu:
Agent commands the agent left running in the background. Clicking one expands it in place to show the command that started it and its live output.
If the command writes its output to a file, that file is followed and named ("Following dev.log").
Each command has its own Stop, which ends only that command.
Terminals in the thread with a running process. Clicking opens the terminal, and Stop sends Ctrl+C.
A command the agent left running, listed under Running. The arrow opens it.
Expanded: the command, the log file it writes to (wsl-app.log), and that file's live output.
Two dev servers Claude started. The Vite one is expanded with its output, and its own Stop shows on hover.
Notes on how it behaves:
Cheap: output is only read while a row is expanded, so a chatty dev server costs nothing when nobody is looking.
Narrow: the server only reads a file named in a command the agent ran in that thread, and only while that command is running.
Older servers: a client connected to an older server keeps today's behavior.
Claude only for now: it's the only provider that reports background tasks to T3 Code today. Others could follow if they report their background work.
Web and desktop: mobile isn't included in this first version.
Implementation
I have a working version as a draft PR: #34592. It adds two RPCs (stop one background task, and subscribe to its output), Claude adapter support using the SDK's stop_task, and the Running section. It also fixes Claude's output location on Windows.
Should agent-started servers on Windows also show up under Servers with their port? Today the port scan can't tell which thread started them on Windows, so they only appear here as running commands. (fix(web): show running servers on thread rows #24606 attributes servers by process ancestry but also skips Windows.)
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Problem
When an agent starts something in the background, such as a dev server, a watcher or a long test run, it's hard to keep track of afterwards:
> dev.log 2>&1, which agents do a lot), there's nothing to see anywhere in T3 Code.Example: I asked Claude to start the backend (.NET API) and frontend (Vite) of another project. It started both in the background, and the only way to see whether they came up was to ask the agent or find the log files myself.
Proposal
A Running section in thread details, on by default and toggled from the existing "Show in thread details" menu:
A command the agent left running, listed under Running. The arrow opens it.
Expanded: the command, the log file it writes to (
wsl-app.log), and that file's live output.Two dev servers Claude started. The Vite one is expanded with its output, and its own Stop shows on hover.
Notes on how it behaves:
Implementation
I have a working version as a draft PR: #34592. It adds two RPCs (stop one background task, and subscribe to its output), Claude adapter support using the SDK's
stop_task, and the Running section. It also fixes Claude's output location on Windows.Questions for maintainers
Related
Environment
Windows desktop app, Claude provider.
All reactions