v0.1.0 — Open source

Turn any MCP Server into a ready-to-go AI Agent

nibchat is a self-hosted chat interface that connects to your tools and services and exposes them as a conversational AI agent. No login, no accounts — just open the browser and start chatting.

How it works

Three simple pieces. Everything else just works.

Step 1

Point to your tools

Tell nibchat where your MCP server is — a database, an API, internal service, or any tool your team uses. One URL in a config file.

Step 2

Deploy in one command

Run docker compose up and nibchat is live. Everything is bundled in a single container — no extra infrastructure needed.

Step 3

Your team starts chatting

No login. No accounts. Users open the browser and instantly have a conversational AI that can actually use your tools and answer in real time.

Your tools & services
(any MCP server)
nibchat (self-hosted)
👤 Your users
(any browser)

Everything you need, nothing you don't

nibchat ships with the features that actually matter for deploying AI to a real team.

Zero friction for users

No login, no accounts, no passwords. Each user gets an anonymous session automatically — just open the browser and start chatting.

Brand it as your own

Set a custom agent name, avatar, description, and starter messages through a single YAML file. Ship a fully branded assistant — not a generic chatbot.

Self-hosted & private

Runs entirely on your own infrastructure. Your conversations, your data. Nothing leaves your environment.

MCP-UI & MCP Apps

Interactive UIs inside the chat

Fully compatible with the MCP-UI and MCP Apps open standards. Tools can return live interactive web UIs — dashboards, forms, data visualizations — rendered as sandboxed iframes right inside the conversation.

Learn about MCP-UI →

Built-in knowledge base (RAG)

Drop PDF, TXT, or Markdown files in a folder and your agent gains instant knowledge. No extra database or infrastructure — everything in SQLite.

Any AI provider

Works with OpenAI, OpenRouter, Ollama, LM Studio, or any OpenAI-compatible API — including local models running entirely on your machine.

Connect multiple tools at once

Point nibchat at several services simultaneously. It discovers all available tools automatically and makes the combined set available to the AI.

Files & images in chat

Users can attach images and PDFs directly in the chat. Files persist across the conversation and can be referenced in follow-up messages.

One command to deploy

Single Docker container. Single docker compose up. The server, UI, database, and file storage are all included.

See it live

These agents are real, running nibchat deployments. Click to try them right now — no login required.

Get started in minutes

You only need Docker. No Node.js, no Python, no build steps.

1

Create a docker-compose.yml

docker-compose.yml
services:
  nibchat:
    image: ghcr.io/cadugrillo/nibchat:latest
    ports:
      - "3000:3000"
    environment:
      - MCP_SERVER_URL=http://your-mcp-server:8080
      - SESSION_SECRET=change-me-to-a-random-string
      # Optional: shared API key for all users
      # - OPENAI_API_KEY=sk-...
    volumes:
      - nibchat-data:/app/data
    restart: unless-stopped
    extra_hosts:
      - "host.docker.internal:host-gateway"

volumes:
  nibchat-data:
2

Start it

$ docker compose up -d
3

Open your browser

 http://localhost:3000

Want the full setup guide, agent config options, and OAuth docs?

Read the full docs on GitHub