MCPFlo is a desktop application for visually testing Model Context Protocol (MCP) servers. Think of it as Bruno or Postman, but for MCP — instead of poking at REST/GraphQL APIs, you connect to an MCP server, browse everything it exposes, and invoke it directly, all through a GUI.

MCP servers expose their capabilities as tools, resources, and prompts, and typically these are consumed by an AI model inside a client like Claude or an agent framework. That’s great for production use, but it makes servers hard to test in isolation — to see what a tool actually returns, or to inspect a raw response, you’d otherwise need to wire up a full AI client or write a throwaway script. MCPFlo removes the model from that loop entirely: you talk to the server directly, with real inputs and real responses, and see exactly what’s happening at the protocol level.

With MCPFlo you can:

• Connect to one or more MCP servers over stdio or Streamable HTTP, including servers that require OAuth 2.1 authentication.
• Browse the full capability tree — every tool, resource, and prompt the server exposes — with search and a token/context-budget estimate.
• Invoke any tool or prompt through an auto-generated, schema-driven form, and inspect the response (JSON, images, audio, embedded resources) exactly as the server sent it.
• Revisit past calls with a lightweight, replayable history.

Who it’s for

• MCP server authors who want fast feedback while building — invoke a tool the moment you add it, without restarting an AI client or writing a test harness.
• Integrators and QA engineers validating that a third-party or in-house MCP server behaves as documented before wiring it into a production agent.
• Anyone debugging an agent that talks to an MCP server, who needs to reproduce a specific tool call outside the agent to see if the problem is in the server or the agent’s usage of it.

It is not an AI chat client and doesn’t involve a model — it’s a protocol-level inspector and testing tool, for humans who need to see and drive MCP servers directly.