Rate limits

The redirect service allows 100 requests per 60 seconds, counted per IP address.

Exceeding it returns 429 with:

{ "error": "Too many requests." }

The limit is checked before anything else, so a rate-limited request never reaches the database and never becomes a redirect.

What counts

Everything. Following a link, loading the index, and requests that turn out to be mistakes all count against the same allowance, because the check happens before the service knows which is which. A script hammering the index with bad names will rate-limit itself just as effectively as one following real links.

Living within it

For people following links this is not a limit you can reach by hand. It exists for scripts.

If you are building something against the index, fetch it once and cache the result rather than calling it per item. The whole list comes back in a single response, so there is rarely a reason to make more than one request: filtering client-side costs you nothing and keeps you well inside the limit.

If you do get a 429, wait for the window to pass rather than retrying immediately. Retries count too, so a tight retry loop keeps the window full and extends the problem.

Shared addresses

The count is per IP, so everyone behind one office connection or VPN egress shares an allowance. It is generous enough that this rarely matters for people clicking links, but it is worth knowing if you deploy something automated inside a network where others are using the same links.