MCP SERVER & CLIENT

Your workspace, on the AI bus.

Rostyman speaks the Model Context Protocol — as both a server and a client. Let AI agents like Claude Desktop browse your collections, run requests, and read responses. Or connect out to any MCP server from within the app.

Why this is different

MCP is how AI agents get real tools instead of just talking. We're one of the first API clients to ship MCP both ways — expose your workspace, or test and debug other people's MCP servers. No plugin. No external proxy. Just check the box.

Rostyman as MCP Server

Your workspace becomes a set of tools Claude Desktop (or any MCP client) can call. Turn it on from Settings. Copy the registration snippet. Your AI now has read and write access to your API collections — with full audit logging.

13 tools

  • List collections, folders, requests
  • Execute any request with params + headers
  • Search across all collections
  • Read and write environment variables
  • Read and write global variables
  • Read vault (with permission prompt)
  • List and run workflows
  • Read request history
  • Read mock servers and their routes
  • Create new requests and collections
  • Export a collection as .rostyman
  • Run a collection runner
  • Get the current selection / active request

4 resources

  • rostyman://collections

    Full collection tree as browsable JSON.

  • rostyman://environments

    All environments and their variables.

  • rostyman://history

    Rolling request history with response summaries.

  • rostyman://vault

    Vault secret keys (values gated by permission).

Three ways to connect

stdio

Claude Desktop registers Rostyman via stdio with a single JSON config entry. Zero network config.

HTTP / SSE

localhost:3100 with auth token, rate limiting, and request logging. Great for web-based clients.

TLS + Tunnels

HTTPS via mkcert for local TLS. Expose remotely with ngrok or cloudflared (auto-detected).

Rostyman as MCP Client

Test other people's MCP servers the way you'd test a REST API. Add a server, list its tools and resources, call them directly, inspect responses. Ideal if you're building MCP servers yourself.

stdio and HTTP transport support
Browse tools and resources by server
Call tools with JSON argument editor
Fetch and preview resources
JSON-RPC request-level logging
Connected clients tabbed UI with disconnect
Status-bar indicator when active
Date-based log files with ring-buffer rotation

Give your AI real tools.

The API client built for the age of AI agents.

Download Free