Use the in-app installer
- Open Settings → Code Harnesses and select the affected harness.
- If a missing-CLI panel appears, choose Install followed by that harness’s name and review the command, download host, and any alias warning.
- Choose Download and run it only when you intend to run that installer. Wait for completion or a displayed failure.
- Read the failure before Try again. Open install docs takes you to the provider’s installation guidance.
On a successful exit, Rudr refreshes executable detection, models, and access. That sequence is not proof that login or a paid first turn succeeded.
Recover with a manual installation
- Follow the native-platform instructions linked from the individual provider guide. Match your shell to the command, especially PowerShell versus CMD on Windows.
- Verify the provider executable in a new terminal:
claude --version,codex --version,grok --version, oropencode --version, as appropriate. Cursor requires the provider-specificcursor-agentexecutable for Rudr; a genericagentcommand is ambiguous. - Fully quit Rudr and launch it again. Return to Code Harnesses and check detection.
- Once access controls appear, complete sign-in or key setup, use Refresh status if offered, and Save settings.
Rudr searches the process PATH and known user/system installation locations. It does not scan the entire disk. An interactive-shell alias, an arbitrary custom directory, or a WSL installation may work in your terminal while remaining unavailable to native Rudr. Reopening Settings alone may not refresh the app’s inherited environment.
Diagnose the remaining failure
If detection still fails, inspect Settings → Usage → Diagnostics. Record OS, Rudr version, provider version, installation method/location, and sanitized error text. Never attach credential files or environment dumps.
If the executable is installed but the route is not ready, move to authentication recovery. Muse has no automated Windows plan and its upstream instructions remain unverified; do not invent a replacement installer.