Skip to main content

Troubleshooting

Find your symptom below (Ctrl+F works). First check for anything odd: curl http://127.0.0.1:8000/api/health tells you whether the server is up at all.

"cptr: command not found"​

The package landed in a different Python environment than the one on your PATH. Install it into the Python you actually run:

python -m pip install cptr

If you installed with pipx or inside a virtualenv, activate that environment first (or use pipx run cptr run / uvx cptr@latest run). Then start it with cptr run.

"Port 8000 already in use"​

Something else owns the port. Either run on another one:

cptr run --port 8001

or find and stop the owner: lsof -i :8000 (macOS/Linux) or netstat -ano | findstr :8000 (Windows).

The setup URL doesn't work / I lost the setup token​

The setup token is printed once when cptr run starts, and the setup URL only works while no account exists yet. If you lost it, just restart cptr run: a fresh token is printed. If an account already exists, the setup URL is dead by design; go to http://localhost:8000 and log in normally.

I forgot my password / I'm locked out​

If another admin exists, they can reset your password in Settings → Admin → Users. Two things that trip people up first:

  • Login is rate-limited to 5 attempts per minute per IP. If you're getting 429, wait a minute and try again.
  • Rotating [server] secret in config.toml signs out every session; it does not reset any password.

If you're the sole admin, stop the server. There is no CLI password reset. Your options, in order:

  1. Restore a backup of app.db from when the password still worked.
  2. Last resort: open app.db with a SQLite client and delete the user rows (the account and auth tables), then restart. First-run setup triggers again and you create a new admin. Your config (config.toml) and workspace chats (<workspace>/.cptr/chats/) survive; only the accounts are recreated.

My phone can't reach it​

localhost on your phone means the phone itself. It will never reach your computer.

  • Same Wi-Fi: restart with cptr run --host 0.0.0.0, open http://<your-computer-ip>:8000 on the phone, and allow the port through the host firewall.
  • Away from home: you need Tailscale or a tunnel; see phone and remote access.
  • Host asleep: nothing can serve a sleeping machine; see keep it running.

Terminal won't open on Windows (VCRUNTIME140.dll)​

The terminal backend needs the Microsoft Visual C++ Redistributable. Install the x64 version from Microsoft, restart your machine, and start cptr run again.

Docker: setup wizard reappeared / my state is gone​

The container started without its /data volume, so it looks like a fresh install. Your data is almost certainly still in the volume; never remove it while diagnosing. Recreate the container with the volume attached:

docker run -d -p 8000:8000 -v cptr-data:/data ghcr.io/open-webui/computer:latest

Your account, workspaces, and settings come back with it.

Docker: SQLite can't write /data​

You bind-mounted a host directory that the container user can't write to. Fix the ownership of the host directory, or (simpler and more robust) use a named volume (-v cptr-data:/data) instead of a bind mount. Don't run the container privileged as a shortcut.

My project folder isn't visible in Docker​

The container only sees what you mount into it. Add your project as a bind mount and use the container path as the workspace path:

docker run -d -p 8000:8000 -v cptr-data:/data -v ~/code/myproject:/workspace/myproject ghcr.io/open-webui/computer:latest

Then add /workspace/myproject as the workspace.

Agent model missing from the selector / detection not ready​

Check the profile status in Settings → Admin → Agents; each status has a specific fix:

  • not found: the agent CLI isn't on the server's PATH. Install it, or put the absolute path in the profile's Command field.
  • missing dependency: Claude Code also needs the claude-agent-sdk Python package in cptr's environment: pip install claude-agent-sdk.
  • auth unknown: run the agent's own login on the host machine (claude, codex login, agent login, cline auth, …).

Detection results are cached for about 30 seconds, so give it a moment after fixing. Full setup per agent: coding agents.

My bot doesn't answer​

Work down this list:

  1. Is the server running, and is the bot's active toggle on in Settings → Admin → Bots?
  2. Did the token pass the verify check when you saved it?
  3. Is your platform user ID actually in Allowed senders? A non-empty list silently ignores everyone else.
  4. Discord and Slack need the websockets Python package: pip install websockets.
  5. WhatsApp is webhook-only; the Meta webhook URL must point at a publicly reachable https://<your-host>/api/webhooks/whatsapp/<bot_id>.

Full setup per platform: messaging bots.

Terminal or chat disconnects behind my reverse proxy​

Computer uses Socket.IO: terminals and streaming die if your proxy doesn't forward WebSocket upgrade headers (Upgrade / Connection). Add them to your proxy config; see reverse proxy for nginx/Caddy/Traefik snippets.

Is the server healthy?​

curl http://127.0.0.1:8000/api/health

Returns {status, uptime_seconds, pid} and needs no login. If this fails on the host itself, the server isn't running; check the terminal where you started cptr run.

This content is for informational purposes only and does not constitute a warranty, guarantee, or contractual commitment. Open WebUI is provided "as is." See your license for applicable terms.