mattcarpenterwkto/simple-enterprise-mcp-chat repository overview

⁠Simple Enterprise MCP Chat

A self-hosted web app for chatting with AI models (OpenAI, Claude, Ollama, LM Studio) that call tools from Workato Enterprise MCP servers — or any MCP server. It includes a side-by-side benchmarking workspace so you can run one prompt across several models or servers and compare token cost, latency, and tool reliability.

Everything runs on your own machine; your conversations, API keys, and sign-ins stay in a local Docker volume and never leave your computer.

⁠Quick start

You need Docker Desktop⁠. Then:

docker run -d \
  -p 127.0.0.1:8501:8501 \
  -p 127.0.0.1:8080:8080 \
  -v "$(pwd)/data:/app/data" \
  --name mcp-chat \
  mattcarpenterwkto/simple-enterprise-mcp-chat:latest

Open http://localhost:8501⁠ in your browser. Your data is saved in a data/ folder in the current directory.

⁠Or with Docker Compose (recommended)

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

services:
  mcp-chat:
    image: mattcarpenterwkto/simple-enterprise-mcp-chat:latest
    ports:
      - "127.0.0.1:8501:8501"   # web UI
      - "127.0.0.1:8080:8080"   # OAuth sign-in callback
    volumes:
      - ./data:/app/data        # your data (DB, tokens, logs) — kept next to this file
    restart: unless-stopped

⁠First-time setup (in the browser)

1. ⚙️ Settings → add an OpenAI and/or Claude API key.
2. 🔌 MCP Servers → add a server (name + URL + auth type).
3. For OAuth servers, click 🔐 Re-authenticate — a sign-in link appears; click it and log in.
4. 💬 Home → pick a model and start chatting.

⁠Configuration

No API keys or server configs are baked into the image — you enter them in the app's UI.

⁠Tags

• latest — most recent release
• vX.Y.Z — specific versions

Built for linux/amd64 and linux/arm64 (Intel/AMD PCs and Apple-Silicon Macs).

⁠Source & docs

GitHub: https://github.com/MattCarpenter-Workato/simple-enterprise-mcp-chat⁠

⁠License

MIT