LibreChat

What is LibreChat?

LibreChat is an enhanced open-source chat interface supporting many AI providers, agents, MCP servers and custom presets. It gives teams one interface across multiple models while keeping conversation history on their own server.

Best for

Teams that want one interface across multiple AI providers

Why choose LibreChat

LibreChat is the answer to the mess of juggling half a dozen AI subscriptions and browser tabs. It is a single ChatGPT-style interface that talks to many providers — hosted APIs, local models, whatever you have keys for — and lets you switch between them mid-conversation, keep per-project presets, and give a household or small team separate accounts with shared history. It is the interface layer you would otherwise build yourself, and it means your conversation history lives somewhere you can back up.

Replaces

• ChatGPT
• Poe
• OpenRouter chat

Key features

• Multiple providers in one interface
• Agents and MCP support
• Custom presets and personas
• Multi-user with conversation history

What to watch out for

Bring your own keys, and bring your own cost model: if you point it at paid APIs, this is a nicer way to spend money, not a way to spend less, and a busy household can run up a real bill. The feature set changes quickly, so guides go stale and some options only exist in recent versions. Running it for multiple users adds genuine administration — accounts, rate limits, access control — that a single-user local setup happily avoids.

How to deploy

• Docker Compose

Getting started

Compose deployment with a real Mongo database, since conversations live there and losing it loses everything. Configure one provider and get a working chat before adding others, so a configuration error is easy to localise. If you intend to give other people accounts, set registration to invite-only and configure access control before announcing it. Point the agents feature at a local model for experimentation, and keep paid API keys on a separate account from anything production so spend is easy to attribute.

Typical setup

A common setup is a small server running LibreChat with MongoDB, connected to a mix of a local model server for private work and API keys for occasional heavy lifting. Households and small teams give each person an account, which turns it into a shared private assistant with a common history. It is usually placed behind authentication on the home network or a VPN, since the conversation archive is as sensitive as anything else on the machine.

Who should look elsewhere

Not a way to save money. If you connect it to commercial APIs you are paying for every message, with more people sharing the bill and less friction spending it. Anyone looking for the best available model quality with no setup should use a hosted assistant; self-hosting here buys privacy and flexibility, not capability.

Project health

• GitHub stars: 45,167
• Last code push: 2026-10-01
• Open issues: 822
• Status: actively developed

Figures pulled from the GitHub API and refreshed periodically.