One gateway. Fleet cost caps, loop detection, dedup cache, audit logs. Drop in front of any LLM API — five lines, two minutes.
4 agents chain a task. The pipeline budget caps at $3.00. When the Writer ↔ Reviewer loop pushes cumulative spend past the cap, the gateway pauses mid-flight — before the bill compounds.
Monthly cap
$5.00
of $5.00 / mo remaining
Pipeline budget
$0.00 / $3.00
client-side · no signup
From a five-line drop-in to fleet-wide audit logs. Each primitive ships independently — pick what you need, ignore the rest.
No decorator, no callback handler, no framework adapter. Point your agent at the gateway URL with a bearer token and you get budget, loop detection, and cache automatically. Works with LangChain, CrewAI, AutoGen, LlamaIndex, or any framework that makes HTTP calls.
Cap how much an entire pipeline can spend — not per-agent, per-fleet. When the ceiling hits, every agent in the pipeline pauses automatically. Monitoring tells you what went wrong; enforcement stops it from continuing.
Five agents in the same pipeline call the same upstream API. The first request hits upstream; the next four are served from cache. Dedup happens across the pipeline, not per-agent — savings stack quietly while your monthly burn drops without code changes.
When two agents enter an infinite back-and-forth, the gateway catches the cycle by tracking call patterns at the pipeline level. The pipeline pauses, a webhook fires, the team wakes up to a flagged session instead of a five-figure invoice.
Every call routed through the gateway is tagged with the originating agent. The dashboard shows which agent burned what — so when something runs hot, you know exactly which agent caused the bottleneck and how much that specific run cost.
If your agents pay for APIs on-chain instead of through API keys, Gatewards implements the open x402 protocol end-to-end — USDC settlement on Base, EIP-3009 transfers, JWT receipts. Independent of proxy mode; opt-in only when you need it.
First call from the pipeline hits upstream. Every other agent making the same call is served from cache — across the whole pipeline, not per-agent.
Point your agent at the gateway URL. Pipeline budget, loop detection, and shared cache happen automatically — your agent code is unchanged. Works with any framework that makes HTTP calls.
// Drop in. No decorator, no callback handler.
// Point your agent at the Gatewards gateway.
const upstream = 'https://api.coingecko.com/api/v3/simple/price?ids=bitcoin'
const data = await fetch(
`${GATEWAY}/api/v1/proxy/${encodeURIComponent(upstream)}`,
{ headers: { Authorization: `Bearer ${process.env.GATEWARDS_API_KEY}` } }
).then(r => r.json())
// Pipeline budget, loop detection, and dedup cache happen
// at the gateway. Your agent code is unchanged.If you sell access to a paid API, Gatewards also implements the open x402 protocol. Set a price, receive USDC on Base, the SDK handles 402 responses + JWT receipts.
// Optional: charge for your API on-chain via x402.
import { createPaymentRequiredMiddleware } from '@gatewards/merchant-sdk'
app.get('/api/data',
createPaymentRequiredMiddleware({
price: '0.10',
wallet: '0xYourWallet',
network: 'base',
gatewayPublicKey: process.env.JWT_SECRET,
}),
(req, res) => res.json({ data: 'premium' })
)Self-host free under MIT, or use the hosted gateway with simple usage caps. Every tier feature ships in the binary — managed plans add infra, dashboard, and support.
Hobby + evaluation
Small team
Large fleets
Self-host · MIT · every tier feature · no caps
Optional: x402 on-chain settlement available on every tier — opt in when you want agents paying for APIs in USDC on Base. 1% router fee on settled volume; tier subscription unaffected.