Self-host LibreChat

AI Chat Interfaces

deployed & timed by usprices checked · Jul 2026

Open-source, multi-provider AI chat platform unifying OpenAI, Anthropic, Google, Azure, and local models behind one UI, with agents, RAG, code interpreter, and MCP tool support.

ChatGPT / yr$240

Self-hosted / yr~$55

You keep$185/yr

Key facts

CategoryAI Chat Interfaces

LicenseMIT

StackTypeScript, React, Node.js, MongoDB

Min RAM2048 MB

Dockeryes

Difficulty

Our recommendation

Choose LibreChat when you want one chat UI in front of many providers — OpenAI, Anthropic, Google, Azure, and local models — with agents, RAG, a code interpreter, and MCP tools built in. It's the most feature-complete option here and cleanly MIT-licensed, at the cost of being the fiddliest to stand up: it wants MongoDB and a stack of `.env` keys. If you only need a clean front end for one local backend, Open WebUI is far less to wire.

What you need

Any VPS with at least 2048 MB of RAM
A domain you control — most self-hosted setups need HTTPS in front of them
About an afternoon — budget time for troubleshooting

Install with Docker Compose

Save this as compose.yml and run docker compose up -d:

# LibreChat — official Docker Compose setup
git clone https://github.com/danny-avila/LibreChat.git
cd LibreChat
cp .env.example .env
# add your provider API keys to .env, then:
docker compose up -d  #  →  http://SERVER_IP:3080

vps picksBest VPS for self-hostingbenchmarked quarterly · re-priced monthly

What you take on

LibreChat's breadth is the draw and the catch — every provider it unifies is one more thing to configure and keep honest:

non-negotiableIt's a real stack, not a single container. LibreChat needs MongoDB alongside it (and a vector store if you want RAG) — budget the 2 GB minimum for the app plus a database to back up, not a one-file deploy.

non-negotiableConfiguration is the work. Multi-provider means a .env full of API keys and a librechat.yaml to shape which models and tools appear — powerful, but plan an evening to wire it correctly the first time.

non-negotiableHosted providers aren't free or private. Point it at OpenAI or Anthropic and you're back to per-token bills and prompts leaving your box — the privacy and cost win only lands when you route it to a local backend.