Local setup

Local mode runs an embedded Letta-compatible backend inside Letta Code. Agents, conversations, memory, and provider connections are stored on your machine.

Local mode controls where Letta Code stores agent state. It does not make model inference local by itself. If you connect a remote provider like OpenAI, Anthropic, Gemini, OpenRouter, Bedrock, or ChatGPT/Codex, prompts still go to that provider. For a fully local loop, connect a local inference provider like Ollama, LM Studio, or llama.cpp.

Connect model providers

Use /connect inside the interactive session to add providers to the local backend:

> /connect

You can also connect providers from the shell:

letta --backend local connect anthropic --api-key "$ANTHROPIC_API_KEY"
letta --backend local connect ollama
letta --backend local connect lmstudio
letta --backend local connect llama-cpp
letta --backend local connect chatgpt

These are common examples, not the full provider list. See Providers for the broader provider reference.

For local OpenAI-compatible servers, pass a base URL:

letta --backend local connect lmstudio --base-url http://127.0.0.1:1234/v1
letta --backend local connect llama-cpp --base-url http://localhost:8080/v1

For slow local inference servers, configure a provider-level timeout:

letta --backend local connect lmstudio --base-url http://127.0.0.1:1234/v1 --timeout 600s

Timeouts are stored per local provider in milliseconds. Pass --no-timeout or --timeout false to disable the provider timeout.

Where local state is stored

By default, local backend state is stored in:

~/.letta/lc-local-backend

Local agent memory is stored as a git repository under:

~/.letta/lc-local-backend/memfs/<agent-id>/memory

Use LETTA_LOCAL_BACKEND_DIR to isolate local state for a project or experiment:

export LETTA_LOCAL_BACKEND_DIR="$PWD/.letta-local"
letta --backend local --new-agent

Add .letta-local/ to .gitignore if you create it inside a repository.

Run local mode headlessly

Headless mode also supports the local backend:

letta --backend local -p "summarize this repository"

Use this for local scripts or one-off automation that should use local agent state.

Troubleshooting

My prompts are still going to a remote provider

Local mode stores state locally, but inference follows the model provider you selected. Switch to an Ollama, LM Studio, or llama.cpp model for local inference.

My local model does not appear in /model

Make sure the local inference server is running, then reconnect the provider:

letta --backend local connect ollama
letta --backend local connect lmstudio --base-url http://127.0.0.1:1234/v1

I want a clean local backend for testing

Set LETTA_LOCAL_BACKEND_DIR to a temporary directory before launching Letta Code:

export LETTA_LOCAL_BACKEND_DIR="$(mktemp -d)"
letta --backend local --new-agent