LIVE · BYOK · MPP 402

See, control, prove, and reuse AI agent work.

The control and receipt layer for AI calls. Anthropic Messages-compatible and OpenAI Chat Completions-compatible routes in front of Anthropic, OpenAI, Google, Groq, Bedrock, OpenRouter. BYOK — your provider bills inference; prxy bills the gateway / control layer.

Every call ships a signed Ed25519 receipt with cost, module chain, policy decision, and four canonical hashes. Outcomes anchor a learning signal. Reviewer-promoted candidates become reusable patterns.

up to ~90%
MCP overhead reduction (tool-heavy; ~33% synthetic average)
12
verified modules
6
providers
$0.05
MPP session call

One base URL for supported routes.

$ swap your provider URL

- ANTHROPIC_BASE_URL=https://api.anthropic.com
+ ANTHROPIC_BASE_URL=https://api.prxy.monster
+ PRXY_KEY=prxy_xxx        # your Anthropic key still does the inference

Pay-per-call via HTTP 402.

Hit /v1/agent/messages with no auth to get a payment challenge. Settle with Authorization: Payment , get the LLM response and itemized Payment-Receipt header. MPP-canonical, RFC 9530 digest-bound, single-use challenge.

$ curl without payment

$ curl -i -X POST https://api.prxy.monster/v1/agent/messages \
    -H "Content-Type: application/json" \
    -d '{"model":"bedrock/us.amazon.nova-micro-v1:0","max_tokens":64,"messages":[{"role":"user","content":"hello"}]}'

HTTP/2 402
content-type: application/problem+json
cache-control: no-store
www-authenticate: Payment id="mpp_...", realm="api.prxy.monster",
                  method="stripe", intent="charge",
                  expires="2026-05-03T...", digest="sha-256=:...:",
                  request="<base64url JSON>"

Toggle the intelligence per request.

mcp-optimizer
up to ~90% less MCP tool overhead on tool-heavy workloads
semantic-cache
embedding-keyed cache
exact-cache
byte-match cache
patterns
Golden Loop injection
compaction-bridge
survives auto-compaction
ipc
infinite context
rehydrator
restore session context
prompt-optimizer
token-density rewrites
cost-guard
per-key USD ceilings
tool-cache
dedupe MCP tool calls
router
cheapest provider routing
guardrails
PII + jailbreak detection

Discovery & API surface

The whole monster.