Agent discovery
Nibgate can make paid content discoverable to AI agents without exposing protected content. Agents get structured metadata, verified source context, price information, and unlock routes.
The goal is not for agents to scrape everything. The goal is for creators to publish structured, verified content metadata that agents can discover and act on safely.
Why agents need a discovery layer
Agents need to answer questions like:
- What content exists?
- Who published it?
- Is the source verified?
- What does it cost to unlock?
- What type of content is it?
- Has it been useful to other users or agents?
- Where can the agent unlock or cite it?
Agent-readable content cards
Verified domain, creator identity, site health, and ownership status.
SourceType, title, summary, tags, price, currency, URL, and preview image.
MetadataViews, unlocks, conversion, feedback, receipts, and freshness.
SignalsUnlock route, payment rail, receipt link, and post-unlock reference.
ActionRelationship to agent registries
Agent identity and reputation standards focus on whether an agent can be identified, trusted, or validated. Nibgate focuses on the content side: verified sources, content quality, unlock events, and feedback tied to real interactions.
Discovery surfaces
Agents can discover content through Explore, public metadata APIs, search filters, tags, content type routes, creator profiles, and future machine-readable manifests.
Machine-readable discovery guide
A plain-language agent discovery guide is available at https://nibgate.xyz/discovery.md. It describes every Nibgate endpoint, the payment flow, and the rating flow — no HTML rendering needed, readable directly by any agent.
Public ledger
The ledger at https://nibgate.xyz/ledger (API: GET /api/hub/ledger) provides a live feed of all views, unlocks, payments, and ratings across sites. Agents can filter by ?domain= for per-site activity. Each entry includes timestamps, wallet addresses, and on-chain tx hashes where applicable.
Hub explore API
GET https://nibgate.xyz/api/hub/explore/content?type=article&sort=trending&limit=50
Accept: application/jsonReturns content with pricing, access policy, creator domain, and reputation signals.
Creator manifests
Agents can also poll /nibgate.json directly from creator sites to discover content with full resource metadata including price, currency, recipient wallet, and access policy.
Purchasing content via x402
Once an agent discovers content, it can purchase and unlock it through the same x402 protocol used by browser wallets — no browser, HTML, or widget needed.
Headless purchase flow
import { GatewayClient } from '@circle-fin/x402-batching/client'
const agent = new GatewayClient({
chain: 'arcTestnet',
privateKey: '0x...',
rpcUrl: 'https://rpc.testnet.arc.io'
})
const result = await agent.pay(
'https://creator.example/api/nibgate/access?path=/writing/my-article',
{
headers: { 'x-nibgate-actor': 'agent' }
}
)
// { data: { ok: true, unlockProof, payment }, formattedAmount: '0.01', transaction: '0x...' }The GatewayClient.pay() method handles the full flow:
- Makes the initial request
- Receives HTTP 402 with
PAYMENT-REQUIREDheader - Signs an EIP-3009
TransferWithAuthorizationusing the agent’s private key - Retries with
Payment-Signatureheader - Returns the unlocked content on success
Agent identification
Agents identify themselves via headers:
| Header | Value | Purpose |
|---|---|---|
x-nibgate-actor | agent | Explicitly declare as agent |
Accept: application/json + x402: true | — | Triggers agent detection |
User-Agent | bot pattern | Auto-detected if matches bot regex |
Agent wallet requirements
- USDC balance on Arc Testnet (chain ID 5042002)
- USDC deposited into Circle Gateway via
agent.deposit('10') - Private key accessible to the agent process
Agent-specific pricing
Creators set different prices for humans vs agents in the resource’s access policy:
const resource = {
access: { humans: 'free', agents: 'paid' }, // agents pay, humans free
price: '0.002',
}Quality over raw indexing
Nibgate should prefer verified, useful, paid content over noisy indexing. Ranking can use verified interactions rather than only clicks or self-reported metadata.