One key for 400+ AI models, a USDC balance, and hard limits your agent can't loosen. Every call leaves a receipt, sealed on Solana every hour.
A way to pay, rules it has to follow, and a history other people can check.
OpenAI-compatible. Change the base URL and keep your code. Claude, GPT, Gemini, Llama and hundreds more, all paid from one USDC balance.
Daily caps, a per-call max, model allowlists, a loop breaker and a pause switch, checked on every call. Changing a limit needs the owner wallet's signature.
Every call writes a chained receipt with totals only. Each hour, all new receipts are rolled into one Merkle root and posted on Solana. Anyone can check one.
Use the SDK you already have. Set the model to reins/auto and Reins picks the cheapest model that can do the job, with fallbacks if one goes down.
from openai import OpenAI client = OpenAI( base_url="https://reins/v1", api_key="rk_live_…", ) client.chat.completions.create( model="reins/auto", messages=[{"role": "user", "content": "…"}], ) # receipt written · header x-reins-receipt
Your agent can spend, but only inside the rules you set. Every change to a limit is signed by the owner wallet, so a jailbroken agent or a leaked key can't raise its own cap.
Every agent has a public page: what it spent, how often it was blocked, how long it's been running. Look one up before you pay it, lend to it, or hire it.
Just created, no history
7 days of sealed receipts
30 days, under 1% blocked
Plus $500 spent on record
90 days, clean history
$REINS launches on Reins' own Raydium LaunchLab platform. Half of every Reins fee buys it back and burns it, every hour, on-chain.
Fee on output tokens
On top of the model's own price.
Buys back and burns $REINS
Hourly market buys, burned on-chain. The $REINS trading fee is burned too.
Runs Reins
Servers, the model credit float, development.