Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Mintplex-Labs/anything-llm/llms.txt

Use this file to discover all available pages before exploring further.

Docker is the recommended and most widely supported way to run AnythingLLM in production. A single docker run command pulls the official image, mounts a local storage folder for persistence, and starts the server on port 3001. The image supports both amd64 and arm64 CPU architectures, works on Linux, macOS, and Windows, and includes everything AnythingLLM needs — no separate database server, no build step, no dependency management required.

​
Requirements

  • Docker v18.03+ on Mac/Windows, or Docker v20.10+ on Linux
  • At least 2 GB of RAM available to the container
  • 10 GB of free disk space (more if storing large document sets or local models)
  • Access to an LLM — a local model via Ollama or an API key from a cloud provider

​
Pull the Image

docker pull mintplexlabs/anythingllm
The latest tag always points to the most recent stable release.

​
Running the Container

export STORAGE_LOCATION=$HOME/anythingllm && \
mkdir -p $STORAGE_LOCATION && \
touch "$STORAGE_LOCATION/.env" && \
docker run -d --rm -p 3001:3001 \
--cap-add SYS_ADMIN \
-v ${STORAGE_LOCATION}:/app/server/storage \
-v ${STORAGE_LOCATION}/.env:/app/server/.env \
-e STORAGE_DIR="/app/server/storage" \
mintplexlabs/anythingllm
After the container starts, open http://localhost:3001 in your browser to complete the setup wizard.

​
Volume Mounts and Data Persistence

The two mounts in the docker run command are essential for keeping your data safe across container restarts and image upgrades:
MountPurpose
${STORAGE_LOCATION}:/app/server/storagePersists your SQLite database, uploaded documents, vector indexes (LanceDB), and locally cached models
${STORAGE_LOCATION}/.env:/app/server/.envPersists your environment configuration so settings survive container recreation
Without these volume mounts, all your data is lost when the container stops or is recreated. Always mount your storage directory to a path on the host machine before running the container for the first time.
If you’re using Docker Compose, the named volume anythingllm_storage achieves the same persistence. Update the device: path under driver_opts to point to the directory on your host where you want data stored.

​
Port Mapping

AnythingLLM’s server listens on port 3001 inside the container. The -p 3001:3001 flag maps that to port 3001 on your host. To use a different host port (e.g. 8080), change the left side of the mapping: 
-p 8080:3001
You can also override the internal port with the SERVER_PORT environment variable — but in most cases the default is fine.

​
Key Environment Variables

The full list of supported variables is in docker/.env.example. The most important ones are:
VariableDefaultDescription
SERVER_PORT3001Port the server listens on inside the container
STORAGE_DIR/app/server/storageAbsolute path to the storage directory inside the container
UID1000User ID the container process runs as
GID1000Group ID the container process runs as
JWT_SECRET(unset)Secret used to sign session tokens — set a long random string in production
AUTH_TOKEN(unset)Single-user password for remote/cloud deployments
DISABLE_TELEMETRYfalseSet to true to opt out of anonymous usage telemetry
LLM_PROVIDER(unset)LLM backend to use (e.g. openai, ollama, anthropic)
VECTOR_DBlancedbVector database backend (e.g. lancedb, pinecone, chroma)
EMBEDDING_ENGINEnativeEmbedding model provider (e.g. native, openai, ollama)
You can set these variables either with -e KEY=VALUE flags on the docker run command, or by writing them to the .env file mounted at /app/server/.env.

​
LLM Provider Examples

# OpenAI
-e LLM_PROVIDER=openai \
-e OPEN_AI_KEY=sk-your-key-here \
-e OPEN_MODEL_PREF=gpt-4o \

# Ollama (running on host machine)
-e LLM_PROVIDER=ollama \
-e OLLAMA_BASE_PATH=http://host.docker.internal:11434 \
-e OLLAMA_MODEL_PREF=llama3.2 \
-e OLLAMA_MODEL_TOKEN_LIMIT=4096 \

# Anthropic
-e LLM_PROVIDER=anthropic \
-e ANTHROPIC_API_KEY=sk-ant-your-key-here \
-e ANTHROPIC_MODEL_PREF=claude-sonnet-4-6 \

​
About UID and GID

The container runs as the user specified by UID and GID (both default to 1000). If your host user has a different UID or GID, the container process may not have permission to read from or write to the mounted storage directory. Check your host UID with id -u and your GID with id -g, and set UID and GID to matching values in your .env file or as -e flags.
# Check your host UID and GID
id -u   # e.g. 1001
id -g   # e.g. 1001

# Pass them to docker run
docker run -d --rm -p 3001:3001 \
  --cap-add SYS_ADMIN \
  -e UID=1001 \
  -e GID=1001 \
  -e STORAGE_DIR="/app/server/storage" \
  -v $HOME/anythingllm:/app/server/storage \
  -v $HOME/anythingllm/.env:/app/server/.env \
  mintplexlabs/anythingllm

​
Connecting to Host Services

If you run a local LLM or vector database on your host machine, use host.docker.internal instead of localhost in URLs: 
# Instead of:   http://localhost:11434
# Use:           http://host.docker.internal:11434
On Linux, add --add-host=host.docker.internal:host-gateway to your docker run command to enable this hostname. Alternatively, use http://172.17.0.1:11434 which is the default Docker bridge gateway on Linux.

​
GPU Support

AnythingLLM itself does not perform GPU inference — the LLM runs in a separate process (e.g. Ollama or LM Studio on the host). If you’re using the built-in local embedder or Whisper transcription and want GPU acceleration, you can pass GPU devices to the container: 
# NVIDIA GPU
docker run -d --rm -p 3001:3001 \
  --cap-add SYS_ADMIN \
  --gpus all \
  -v $HOME/anythingllm:/app/server/storage \
  -v $HOME/anythingllm/.env:/app/server/.env \
  -e STORAGE_DIR="/app/server/storage" \
  mintplexlabs/anythingllm
The --gpus all flag requires NVIDIA Container Toolkit to be installed on the host.

​
Building from Source

For development or custom builds, you can build the image locally: 
git clone https://github.com/Mintplex-Labs/anything-llm.git
cd anything-llm/docker
cp .env.example .env
# Edit .env as needed, then:
docker-compose up -d --build
This builds the full image from source and starts the container. The build takes a few minutes on first run.

​
Troubleshooting

API not working / can’t log in on a remote server? If AnythingLLM is running on a remote machine (e.g. EC2 at 192.168.1.222), you need to set VITE_API_BASE before building: 
# frontend/.env.production
VITE_API_BASE="http://192.168.1.222:3001/api"
Then rebuild with docker-compose up -d --build. Ollama errors (ECONNREFUSED)? Make sure you’re using http://host.docker.internal:11434 as the Ollama base path — not localhost or 127.0.0.1. On Linux, ensure --add-host=host.docker.internal:host-gateway is in your docker run command. Need help? Join the Discord community.

Build docs developers (and LLMs) love

Get started for freeTalk to us