Quickstart
Learn the basics of Letta Code
Setting up your agent
Section titled “Setting up your agent”If you’re using Letta Code for the first time, run the /init command to initialize the agent’s memory:
> /initThis prompts your agent to research your project, ask about your preferences, and organize its memory blocks. The agent can do either standard (quick) or deep research (comprehensive). See the Memory guide for full details on what /init does.
Over time, the agent will update its memory as it learns. To actively guide your agent’s memory, use the /remember command:
> /remember [optional instructions on what to remember]Connecting to an existing Letta agent
Section titled “Connecting to an existing Letta agent”Letta users may have an existing agent designed for other purposes, such as writing or personal assistants. Letta Code is a general-purpose interface to any agent on your server.
To connect to an existing agent, use the --agent flag:
letta --agent [existing-agent-id]If the agent wasn’t created inside of Letta Code, it won’t have access to coding tools (which allow it to connect to your computer) by default. Use the /toolset command to switch toolsets:
> /toolsetThis opens a toolset selector where you can choose between different tool configurations (default, codex, gemini). Select the appropriate toolset for your use case.
Running Letta Code in headless mode
Section titled “Running Letta Code in headless mode”You can also run Letta Code in headless mode, which allows you to compose your agents as UNIX programs.
letta -p "Look around this repo and write a README.md documenting it at the root level"Next steps
Section titled “Next steps”Read more about Letta Code’s rich feature set:
- Skills - Create reusable modules to extend your agent
- Memory - Understand the hierarchical memory system
- Headless mode - Run Letta Code non-interactively
- Commands reference - All CLI and interactive commands