Labsco
MCP SERVER

RequestRepo MCP

by hofill

Run a requestrepo catch-all endpoint from the chat — wait for a callback, read what arrived, change what it serves.

API Specifications & HTTP Tooling
Summary
wait_for_request is what turns a callback into a step in the conversation.

Testing a webhook usually means switching to a browser tab and refreshing until something shows up. Here the agent blocks on wait_for_request and then reads what arrived. Every mutation needs confirm=true, so nothing rewrites your DNS records or served files on a first attempt.

What it is

A Python server over requestrepo, the request-capture service. It covers the whole surface: captured requests, sharing, DNS records and served files.

What you get
  • wait_for_request to block until a callback lands, list_requests for what has already arrived
  • delete_request and delete_all_requests to clear the log
  • share_request and get_shared_request to hand a captured request to someone else
  • add_dns, list_dns, update_dns and remove_dns for the DNS records on your subdomain
  • set_file, get_file, list_files and update_files for what the endpoint serves back
  • session_info for the current session
  • Mutating tools gated behind confirm=true
  • Binary fields returned in a JSON-safe bytes envelope
  • Stdio by default, streamable-http optional
Requirements

Python, installed from the repository. Everything is optional: with no REQUESTREPO_TOKEN the server creates a new session for you, and an admin token can be supplied instead. The host it talks to, the port and protocol, whether certificates are verified, the default wait before a timeout and the largest response it will return all have defaults, and all can be overridden in the environment.

Setup effort

Build from source — clone the repository and build it, then point your client at the binary