Overview

ChatMCP is an open-source, cross-platform graphical client for managing Model Context Protocol (MCP) server connections and chatting with selectable large language models (LLMs). It provides a compact GUI that centralizes connections to one or more MCP-compatible servers, lets you pick models, start chat sessions, and manage conversation history — useful when you switch between cloud-hosted or local LLMs during development and experimentation.

For developers, ChatMCP simplifies routine tasks such as testing model responses, comparing behavior across models, and debugging integrations that use MCP-style APIs. The app is available for Linux, macOS, and Windows and is maintained in the repository: https://github.com/AI-QL/chat-mcp.

Features

Cross-platform desktop application (Linux, macOS, Windows)
Manage multiple MCP server connections (URL, auth/token, metadata)
Select from available models on a connected server and switch models mid-conversation
Interactive chat UI with streaming responses and message history
Conversation export/import (JSON or plain text)
Per-connection settings: default model, context window size, and system prompts
Simple diagnostics: ping servers, view available models and server info
Open-source codebase so you can extend or self-host the client

Installation / Configuration

Two primary ways to get started: download a prebuilt release or build from source.

Download a released binary

Visit the repository releases: https://github.com/AI-QL/chat-mcp/releases
Choose the binary/installer for your platform (Linux/macOS/Windows) and follow the platform-specific installer steps.

Build and run from source (common workflow)

Clone the repository and run the typical Node/Electron development commands. Check the repo README for exact dependency requirements.

# clone the repo
git clone https://github.com/AI-QL/chat-mcp.git
cd chat-mcp

# install dependencies (example using npm)
npm install

# run in development mode
npm run start

# build distributables (platform-specific)
npm run build

Configuration examples ChatMCP stores connection settings per user. You can create or edit a JSON configuration file to add MCP servers:

{
  "connections": [
    {
      "name": "Local-Llama",
      "url": "http://localhost:8080",
      "authToken": "my-local-token",
      "defaultModel": "llama-3b",
      "contextWindow": 2048
    },
    {
      "name": "Cloud-Provider",
      "url": "https://mcp.example.com",
      "authToken": "xxxxxx",
      "defaultModel": "gptx-4",
      "contextWindow": 8192
    }
  ]
}

Environment variables (optional) — ChatMCP may read standard env vars for development/testing:

# Example environment variables for a local dev run
export MCP_URL="http://localhost:8080"
export MCP_TOKEN="my-local-token"
npm run start

Available Resources

Repository: https://github.com/AI-QL/chat-mcp — source, releases, and issues
Release assets: prebuilt installers and portable binaries for each platform
Configuration format: JSON or internal settings UI (per-connection)
Diagnostic tools within the GUI: server ping, model list viewer, error/log console
Export formats: conversation JSON, plain text transcript

Note: The exact file locations and configuration schema may vary between releases; check the versioned docs in the repo for specifics.

Use Cases

Model comparison and prompt engineering

Connect ChatMCP to two MCP servers (e.g., a local LLM and a cloud provider). Send identical prompts and compare outputs side-by-side to iterate on prompts and system messages.

Local development and integration testing

Run your MCP-compliant server implementation locally. Use ChatMCP to exercise endpoints and validate streaming behavior, content formatting, or token-limiting behavior without writing custom client code.

Rapid prototyping of chat interfaces

Use ChatMCP to prototype conversation flows and system prompt adjustments. Export conversation transcripts to share with teammates or to embed as test cases for automated evaluation.

Offline or secure environments

Deploy a private MCP server inside a secure network and point ChatMCP to it. Developers can debug model behavior without exposing data to third-party services.

Troubleshooting and diagnostics

When an application reports unexpected model responses or API errors, connect ChatMCP to the same server/credentials and reproduce the request to isolate whether the issue is on the server, model, or client side.

Tips for Developers

Keep connection tokens secure: prefer platform credential stores or environment variables for sensitive tokens when possible.
Use conversation export to create reproducible test cases for model behavior regression tests.
If your MCP server supports model metadata, ChatMCP can display model capabilities (context size, streaming support); use this to set sensible defaults per connection.
For headless CI testing or automation, rely on your MCP server’s REST API directly — the GUI is primarily for interactive exploration and debugging.

If you need platform-specific build instructions or run into issues, open an issue in the project’s GitHub repository and include your OS, ChatMCP version, and any relevant logs.

ChatMCP: MCP Server GUI for Linux macOS Windows