Agent payments
prxy.monster exposes a dedicated Machine Payments Protocol surface for agents that can respond to HTTP 402 Payment Required challenges.
Agent payments are in private preview. The gateway can publish merchant discovery and issue payment challenges. Full payment-token redemption is disabled until the charging path is enabled. API-key access through POST /v1/messages remains the production path today.
Discovery
Wallet-aware agents discover paid routes by reading:
GET https://api.prxy.monster/.well-known/mppThe endpoint returns 404 unless MPP is explicitly enabled for the deployment. When enabled, it advertises the dedicated agent route:
{
"spec_version": "0.1",
"realm": "api.prxy.monster",
"routes": [
{
"path": "/v1/agent/messages",
"method": "POST",
"price_usd_per_call": 0.05,
"category": "llm-generation"
}
]
}Regular API-key endpoints such as /v1/messages and /v1/chat/completions are intentionally not advertised as MPP routes. They require Authorization: Bearer <prxy_key>.
Agent Route
POST https://api.prxy.monster/v1/agent/messagesThe request body uses the same Anthropic-compatible Messages shape as POST /v1/messages. The authentication flow is different:
| Request | Response |
|---|---|
No Authorization: Payment ... header | 402 Payment Required with WWW-Authenticate: Payment ... |
Authorization: Payment ... header present | Private-preview response until payment-token redemption is enabled |
Example Challenge
request
curl -i https://api.prxy.monster/v1/agent/messages \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-4-6",
"max_tokens": 256,
"messages": [{ "role": "user", "content": "Say hi" }]
}'Why a Separate Route?
MPP wallets and API-key clients have different auth shapes. Keeping /v1/agent/messages separate from /v1/messages prevents a wallet retry from colliding with bearer-token auth, and it lets production API-key traffic stay boring.
Production Checklist
Before this becomes a general-availability payment path, the deployment needs:
- Payment-token verification.
- Idempotent charge creation.
- Receipts tied to the provider call result.
- Refund handling when the provider call fails after payment.
- Per-agent and per-wallet spend controls.
- Public docs that say “agents can pay” only after the above is live.