MCP Workbench – Visual Debugger for MCP Servers

TL;DR

TL;DR: MCP Workbench is a free, browser-based visual debugger for Model Context Protocol servers — inspect JSON-RPC traffic, test tools, and explore resources in real time with zero installation.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: mcp-workbench.ai — verified via direct fetch
• Author / organization: Orkes, Inc. — verified via og:site_name and JSON-LD
• HN launch thread: news.ycombinator.com/item?id=46532597 — Show HN by opiniateddev
• License: No explicit license stated on the project page or HN thread (not confirmed open-source)
• Source last checked: 2026-07-17

What Is MCP Workbench?

Model Context Protocol (MCP) is becoming the standard way to connect AI agents to external tools and data sources — but debugging an MCP server has historically meant staring at raw JSON-RPC traffic or writing throwaway test scripts. MCP Workbench (by Orkes, Inc.) is a free, browser-based visual debugger that changes that.

From the official description:

Visual debugger and test client for Model Context Protocol servers. Debug, test tools, and inspect JSON-RPC traffic in real time.

It is a single-page web application — open the URL, point it at your MCP server, and start inspecting. No installation, no CLI, no configuration files.

Key Features

Real-time JSON-RPC inspection Every request and response between your client and the MCP server is captured and displayed in a readable format. You see the full payload — method names, parameters, results, and errors — as they happen.

Tool testing Manually invoke any tool defined by the MCP server. Supply parameters in a form view or raw JSON and hit send. The response appears instantly, making it easy to verify server behavior without a full integration.

Resource browser MCP servers expose resources (static data the agent can read). Workbench surfaces these in a sidebar so you can inspect what data is available before writing an agent prompt.

OAuth support If your MCP server requires authentication, Workbench handles the OAuth flow so you can debug protected endpoints without writing an auth handler.

Getting Started

MCP Workbench is entirely browser-based. The typical workflow:

1. Open https://www.mcp-workbench.ai/ in your browser
2. Enter your MCP server URL (the server must expose the MCP protocol over Streamable HTTP)
3. The connection panel shows connection status and lists available tools and resources
4. Select a tool, fill in parameters, and send a test request
5. Inspect the JSON-RPC response in the traffic panel

No account or API key is required for the basic debugger functionality.

Under the Hood

• Architecture: Single-page application running entirely client-side. No telemetry or data leaves your browser unless you explicitly connect to a remote MCP server.
• Protocol support: MCP over Streamable HTTP — the current recommended transport for MCP servers.
• Authentication: OAuth 2.0 flow for servers that require it.
• Organization: Built by Orkes, Inc., the company behind the Conductor orchestration engine. Orkes also maintains a commercial managed MCP hosting platform alongside this free tool.

FAQ

Q: Is MCP Workbench open source? A: The project page does not explicitly state a license. It is free to use but may not be open source. Check the project page for the latest licensing information.

Q: Do I need to install anything to use it? A: No. It runs entirely in your browser. You only need a running MCP server to connect to.

Q: What MCP transport does it support? A: Streamable HTTP, which is the current recommended transport for MCP.

Q: Can I debug any MCP server, including self-hosted ones? A: Yes — as long as your server exposes an HTTP endpoint for the MCP protocol, you can point Workbench at it.

Q: Does it support authentication? A: Yes — Workbench includes OAuth 2.0 support for MCP servers that require it.

Conclusion

MCP Workbench fills a practical gap in the MCP ecosystem: there was no dedicated visual tool for inspecting what an MCP server actually does. For developers building AI agents that call MCP tools, Workbench makes the debug loop dramatically faster — no more writing test scripts or reading raw JSON. It is free, requires no setup, and runs in any modern browser. If you are working with MCP servers, it is worth keeping open in a tab.