Configuring the open source Docker image

The open source Docker image is not an officially supported product. For questions and support regarding the open source Docker image, we recommend our Discord server to access community support.

Letta Code can connect to a Letta API proxy server. You can run a Letta API proxy server using the open source Docker image.

Prerequisites

Before connecting Letta Code, you need to have a running Letta API proxy service. See the Docker setup guide for detailed instructions.

This example command configures a Letta API proxy server with an OpenAI API key and an Anthropic API key (for Claude models):

docker run \
  -v ~/.letta/.persist/pgdata:/var/lib/postgresql/data \
  -p 8283:8283 \
  -e OPENAI_API_KEY="your_openai_api_key" \
  -e LETTA_MEMFS_SERVICE_URL=local \
  -e ANTHROPIC_API_KEY="your_anthropic_api_key" \
  -e OLLAMA_BASE_URL="http://host.docker.internal:11434/v1" \
  letta/letta:latest

Connecting Letta Code to your proxy server

Set the LETTA_BASE_URL environment variable to point to your server:

export LETTA_BASE_URL="http://localhost:8283"
export LETTA_MEMFS_BASE_URL="http://localhost:8283"
export LETTA_MEMFS_LOCAL=1

Then run Letta Code normally:

letta

You can also set it inline:

LETTA_BASE_URL="http://localhost:8283" LETTA_MEMFS_BASE_URL="https://localhost:8283" LETTA_MEMFS_LOCAL=1 letta

Swapping between models

Models available on your proxy server will be shown in /model. The different LLM API keys you provided on startup will limit the amount of models shown in the model selector.

Note that Letta Code is only tested with the open source Docker image using a limited set of popular models from OpenAI and Anthropic. Attempting to use Letta Code with models not listed in the official models.json file may result in unexpected behavior or bugs.

For unofficial community support, please visit our Discord server.

Using Letta Code with local LLM inference

Although Letta Code can technically be run with any open weights model (e.g. via Ollama), it is a very demanding agent harness, and you are unlikely to get good performance with most open weights models outside of the very best ones (e.g. GLM-4.7, MiniMax M2.1).

We strongly recommend new Letta Code users first experiment with frontier models (e.g. Claude Opus 4.6 or GPT-5.3) to understand Letta Code’s capabilities before experimenting with open weights models.

To configure Letta Code to point at a local Ollama server, use the OLLAMA_BASE_URL when starting the server:

docker run \
  -v ~/.letta/.persist/pgdata:/var/lib/postgresql/data \
  -p 8283:8283 \
  -e LETTA_MEMFS_SERVICE_URL=local \
  -e OLLAMA_BASE_URL="http://host.docker.internal:11434/v1" \
  letta/letta:latest

See our full Docker model provider page for instructions on LM Studio.