Connect your agent to RideMyPark
The MCP server gives Claude, ChatGPT or Cursor what they need to find a spot, read a listing and, if you allow it, pull up your bookmarks. Read-only, from the first tool to the last.
What it is
A Model Context Protocol server over Streamable HTTP. Your agent adds it as a connector, then calls named tools instead of guessing from a web page.
The public catalogue answers with no account: spot search, listings, shops, events, blog, counters. For your own data, the agent goes through OAuth and asks for your consent on screen.
Connecting
One address for everyone. Paste it wherever your agent adds connectors.
- Endpoint
- https://mcp.ridemypark.com/mcp
The first time a tool touches your data, your agent sends you to an authorisation screen. You see who is asking, what it will read, and you can say no.
The tools, by tier
- Public tier
- Everything already public on the site: the spot map, shops, events, the blog and the counters. No account, no token.
- Your identity
- Who you are on RideMyPark: display name, handle, language, public roles.
- Your data
- What you put on the site: bookmarks, lists, submitted listings, agenda, notifications, activity.
What an agent can ask for
Three permissions, plus one to stay connected. You see them all on the authorisation screen, in your language.
What the server cannot do
- No writes.
- There is no tool that creates, changes or deletes anything. A manipulated agent cannot break a thing.
- No token passed on.
- Your agent's token stops at the MCP server. It is never forwarded to the API, which works with its own service identity.
- Revocable in a second.
- You cut an agent from your settings: the consent is gone, and an access token still in flight expires within ten minutes.
- Every call is logged.
- Tool, status, duration: the trace is kept for 90 days. Never the raw arguments, only their fingerprint.