tuui

TUUI is a desktop MCP client that enables cross-vendor LLM API orchestration by exposing MCP server tools, prompts, and resources in a unified UI. It leverages the MCP architecture to let you connect multiple backends and toolchains through a single interface, discover available MCP servers, and configure how they are used in conversations. The built-in features include Tools for exposing actionable capabilities to the model, Prompts and Resources for guiding behavior and data sources, and client-side features like Sampling and Elicitation to steer responses. You can also discover servers in real time via the MCP Registry and use MCP Bundles (MCPB) to bundle toolsets for easy reuse. For remote deployments, TUUI supports Cloudflare-backed remote MCP servers via the mcp-remote workflow, enabling authenticated, cloud-hosted toolchains alongside local ones. In practice, you can configure multiple chatbots or tool providers within the UI, switch between them, and orchestrate calls across vendors during a single chat session.

Prerequisites:

• A supported desktop environment (Windows, macOS, or Linux)
• Node.js installed if you plan to build or run from source (recommended for developers)
• Internet access to download dependencies or MCP server bundles

Installation steps (from source):

# 1) Clone the repository
git clone https://github.com/AI-QL/tuui.git
cd tuui

# 2) Install dependencies (example using npm)
npm install

# 3) Build or run in development mode
npm run build   # or npm run dev

If you prefer using prebuilt releases, download the latest release for your platform from the TUUI Releases page and follow the platform-specific instructions provided there.

Configuration:

• Locate or create the mcp.json (or use the embedded defaults under src/main/assets/config/mcp.json) to define MCP servers.
• Update any server endpoints, API keys, and models as needed.

Usage:

• Launch the application after installation and navigate to the MCP configuration section to add or edit MCP servers.
• Save changes and restart the client to apply new MCP server configurations.

Tips and notes:

• MCP servers may require authentication; ensure OAuth or API keys are configured in your environment or within the mcp.json configuration.
• When using remote MCP servers (e.g., Cloudflare/mcp-remote), ensure the domain is trusted and CORS/auth settings are correctly configured to avoid HTTP 400 or redirect issues.
• If you encounter issues with authentication callbacks, clearing the browser cache on the authentication page can help resolve OAuth-related delays.
• Environment variables commonly used: MCP_ENDPOINTS, MCP_API_KEY, MCP_MODEL_LIST. Refer to your chosen MCP server’s documentation for specifics.
• For testing, you can start with a Cloudflare remote server entry as shown in the example and progressively add local MCP servers (e.g., a local npx-based mcp-remote or other tooling) to compare behavior.