Skip to main content
Applies to LLM and MCP. An oversized body reaches the gateway before anything can judge it: it has to be read, buffered and parsed. Whether it came from a client that attached a PDF to a prompt by mistake or from someone probing for a way to exhaust memory, the cheapest defence is to refuse it by size before reading it.

​
What it checks

The body as received, against two ceilings, in this order: Bytes are checked first. A body that passes the byte ceiling is then counted in characters, so a request can be small in bytes and still refused for length, and a multi-byte script can hit the byte ceiling before the character one.

​
What the client gets

A refused request is answered before any provider is contacted: Requiring Content-Length refuses chunked uploads and clients that stream a body of unknown size. That is the point — a declared size can be judged before the body is read — but it also refuses well-behaved clients that stream, so turn it on only where every caller is yours. In Observe mode the request is measured, recorded on the event as over the limit, and passed through.

​
Where to apply it

Run it gateway-wide as a ceiling, and target a tighter one at applications whose traffic should be small. On a targeted policy the limit covers all of the application’s traffic, on both planes. It is not a substitute for model limits: prompts that do not fit the resolved model are already rejected on the chat path. This policy is about not reading the body at all.