tester -client

The Tester MCP Client is a Node.js-based client that connects to an MCP server over HTTP streamable transport and presents a chat-like UI for interacting with tools exposed by the MCP server. When you run the client, it establishes a stream to the MCP server, forwards your queries, and receives real-time streamed responses that may include tool usage blocks and results. This setup lets you test MCP servers by issuing natural language questions or prompts and observing how the server selects and executes tools to fulfill the request. The client UI (served by the actor) renders the conversation and any tool call traces to give you an interactive debugging surface.

To use it, run the client locally or on your hosting environment, then open the provided UI URL in your browser. The client can be pointed at any MCP server that supports HTTP streamable. You’ll typically need to provide an API key or token via environment variables (for example APIFY_TOKEN) and optionally your LLM provider API key if you plan to use a self-hosted model. Check the logs for a URL to the Tester MCP Client UI, where you can type questions like “What are the most popular Actors for social media scraping?” and observe the dynamic tool usage and responses from the MCP server.

Prerequisites:

• Node.js (14+ or newer) and npm installed on your system
• Access to an MCP server that supports HTTP streamable (e.g., Apify MCP Server)

Installation steps:

1. Clone the repository or install the npm package for the tester MCP client git clone https://github.com/apify/tester-mcp-client.git cd tester-mcp-client

2. Install dependencies npm install

3. Create a .env file or set environment variables (example below) APIFY_TOKEN=YOUR_APIFY_TOKEN LLM_PROVIDER_API_KEY=YOUR_API_KEY

4. Start the client npm start

5. Open the UI URL provided in the logs, typically something like http://localhost:3000, to interact with an MCP server.

Tips and common issues:

• Ensure the MCP server URL you intend to connect to is accessible from your environment and supports HTTP streamable transport.
• If you’re using your own LLM provider, set LLM_PROVIDER_API_KEY in the environment; otherwise, you can rely on the default provider if configured.
• The client UI is served from the public directory of the project; if you modify the frontend, rebuild or re-run as needed.
• In environments with firewalls or proxies, confirm that SSE (server-sent events) and HTTP streaming are permitted between the client and MCP server.
• Check logs for the exact UI URL each run; it may differ per deployment.
• If you need to test against multiple servers, you can run separate instances with distinct mcpServers configurations.