Troubleshooting

Run the built-in diagnostic

Start with /doctor when memory structure or context usage looks unhealthy:

> /doctor

Use /status and /context to inspect the current agent, model, conversation, and context usage.

I cannot find an agent on another computer

Confirm that you signed in with the same Letta account on both computers. Agents created without an account are stored only on the machine where they were created and are not backed up to the cloud.

See Local setup for local storage paths and backup guidance.

A computer is unavailable in the web app

The desktop app or letta server process must remain running on that computer. Check its network connection, then restart the process if the environment does not reconnect.

See Bring your own machine for setup and connection details.

A command is waiting for approval

The Letta app and CLI ask before tools perform actions that are not covered by an allow rule. Review the pending action in the interface, or update your permission settings when the action should be allowed consistently.

See Permissions.

A model or provider is missing

Run /connect to add or refresh a provider, then use /model to choose one of its models.

> /connect
> /model

For local providers, confirm that Ollama, LM Studio, or llama.cpp is running before reconnecting. See Local models for local and cloud-hosted setup options.

Local state or inference is not where I expected

Local setup controls where agent state is stored; it does not make model inference local by itself. Prompts still go to the selected provider unless you connect a local inference provider.

See Local setup.

Get help

If the problem persists, open an issue in the Letta Code repository or ask in the Letta Discord. Include the Letta Code version, operating system, interface, and the exact error message.