DOCUMENTATION

prxy.monster

A composable AI gateway. One BASE_URL away from caching, cost guards, smart routing, and persistent memory across every model your app talks to.

prxy.monster v1 is in early access. The 6 modules below are production-ready. The remaining 6 ship in v1.1 — clearly marked as such throughout these docs.

Get started in 90 seconds

Cloud quickstart

Local quickstart

One docker run. Nothing leaves your machine.

Customize a pipeline

Pick which modules run and in what order.

What you get out of the box

Drop-in compatibility — swap one env var, keep your existing Anthropic or OpenAI client code.
Composable middleware — caching, cost limits, prompt optimization, persistent memory.
Two backends, one API — run on our cloud or on your own hardware.
No token markup — bring your own provider key, pay our gateway tier (or zero, on local).

Browse by topic

Concepts

Pipeline, modules, storage adapters — the mental model.

Modules

Reference for every built-in module + its config.

API reference

Every endpoint. Headers. Schemas. Examples.

Recipes

Pre-built pipelines for common use cases.

Build a module

SDK docs — write your own middleware.

Migrating?

Diffs from OpenRouter, Portkey, Helicone, LiteLLM.

The shortest possible “hello”


export ANTHROPIC_BASE_URL=https://api.prxy.monster
export ANTHROPIC_API_KEY=prxy_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
 
curl https://api.prxy.monster/v1/messages \
  -H "Authorization: Bearer $ANTHROPIC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Say hi"}]
  }'

That request just ran through:

mcp-optimizer — pruned irrelevant tools (none, in this case)
semantic-cache — no hit, continued
Provider call to Anthropic
patterns — looked at the response for fixes worth saving

You can change that pipeline at any time.

Next: cloud quickstart →