Satelink

AI-Agent & Machine-Commerce Infrastructure

Infrastructure for AI agents that discover and use digital services.

Satelink is programmatic infrastructure for AI agents and automated systems to discover services, evaluate options, and access APIs through controlled, machine-readable workflows. Built by Jakuraa.

Paper trading in controlled beta. Live execution is not yet available.

Explore documentation Request beta access

Agent procurement lifecycle

The 7-Step Agent Procurement Lifecycle

Discover, evaluate, authorize, access, meter, pay, settle: Every software agent operates through controlled, machine-readable steps.

01Task Inception
Step 1/7

Task Decomposition & Constraint Analysis

The autonomous agent breaks down user instructions into required computational services and spending bounds.

When an AI agent receives an autonomous goal, it analyzes required inputs, query parameters, latency requirements, and allocated budget ceilings before initiating any network interaction.

  • Deconstructs task into discrete API dependencies
  • Evaluates deterministic spending limits and loss collars
  • Validates schema arguments before dispatch
agent_intent.jsonTask Inception
{
  "agent_id": "agent_alpha_09",
  "objective": "ingest_latest_polygon_state",
  "required_services": [
    "eth_blockNumber",
    "eth_getBlockByNumber"
  ],
  "budget_ceiling_usd": 0.05,
  "max_latency_ms": 1200,
  "fallback_rail": "x402_base"
}
Action TypeAutonomous Intent
Budget LimitStrict Cap (Enforced)
02Service Discovery
Step 2/7

Machine-Readable Service Catalog Query

The agent queries machine catalogues and MCP endpoints to locate compatible digital service providers.

Instead of scraping human documentation, the agent fetches machine-readable manifests (such as /.well-known/satelink.json) or queries Satelink Model Context Protocol (MCP) servers to discover endpoints, schemas, and terms.

  • HTTP GET to standard /.well-known/satelink.json manifest
  • Reads available endpoints, methods, and protocol versions
  • Inspects pricing.json for immutable unit rates
GET /.well-known/satelink.jsonService Discovery
GET https://rpc.satelink.network/.well-known/satelink.json HTTP/1.1
Host: rpc.satelink.network
User-Agent: satelink-agent/1.0

HTTP/1.1 200 OK
Content-Type: application/json

{
  "version": "1.0.0",
  "catalog": {
    "rpc": "https://rpc.satelink.network/rpc/polygon",
    "pricing": "https://satelink.network/pricing.json",
    "auth_schemes": ["api_key", "x402"]
  }
}
StandardRFC 8615
ManifestActive JSON
03Economic Comparison
Step 3/7

Unit Cost, Latency & Quota Verification

The agent checks unit economics and authorization options against execution limits.

Software agents verify cost per call against allocated balances. Satelink provides transparent pricing: metered Polygon PoS RPC at a flat rate per call, with clear keyless x402 bundle options.

  • Flat per-call rate for Polygon PoS JSON-RPC
  • Keyless micropayments via x402 per-call protocol
  • Zero hidden surge charges or dynamic markups
pricing_evaluation.jsonEconomic Comparison
{
  "service": "polygon_rpc_gateway",
  "chain_id": 137,
  "unit_cost_mode": "flat_per_call",
  "payment_options": [
    { "rail": "usdt_prepaid_vault", "network": "polygon_pos" },
    { "rail": "x402_keyless_usdc", "network": "base_l2" }
  ],
  "quote_validity": "deterministic_flat_rate"
}
Pricing ModeDeterministic Flat
SettlementOn-Chain 137
04Route Selection
Step 4/7

Target Route & Protocol Handshake Preparation

The agent selects the endpoint and prepares payload parameters conforming to JSON-RPC 2.0 standards.

With requirements validated and costs checked, the agent prepares the payload. It formats the JSON-RPC request and prepares either its API key or its wallet signer for keyless settlement.

  • Formats standard JSON-RPC 2.0 method payload
  • Verifies gateway health status and DNS resolution
  • Selects preferred payment rail based on account balance
prepared_payload.jsonRoute Selection
{
  "jsonrpc": "2.0",
  "id": 1042,
  "method": "eth_getBlockByNumber",
  "params": ["latest", false],
  "meta": {
    "target": "https://rpc.satelink.network/rpc/polygon",
    "rail_selected": "x402_base"
  }
}
ProtocolJSON-RPC 2.0
Health CheckPassing
05Programmatic Authorization
Step 5/7

Keyless HTTP 402 Challenge or Scoped Key Authorization

Access is authorized keylessly via HTTP 402 on Base (USDC) or via prepaid USDT credit accounts.

If calling keylessly without an API key, the gateway answers with HTTP 402 Payment Required specifying payment terms. The agent signs an authorization header in USDC on Base, or supplies an authorized API key funded with USDT.

  • HTTP 402 handshake with machine-readable payment terms
  • EIP-712 payment authorization signed in USDC on Base
  • Prepaid USDT credits verified atomically via RevenueVault
HTTP 402 Challenge HandshakeProgrammatic Authorization
HTTP/1.1 402 Payment Required
WWW-Authenticate: x402
Content-Type: application/json

{
  "scheme": "x402",
  "network": "eip155:8453",
  "asset": "USDC",
  "amount": "0.10",
  "recipient": "0x966E...7Ad4",
  "facilitator": "https://x402.satelink.network"
}
Status Code402 Payment Required
Keyless RailBase (Chain 8453)
06Workload Execution
Step 6/7

Gateway Execution & Telemetry Streaming

The authenticated request routes through production RPC proxies, streaming the workload result.

The request passes deterministic control gates. The gateway dispatches the workload directly to dedicated Polygon PoS nodes, streaming back the complete, verified JSON-RPC response with low latency.

  • Strict input sanitization and batch size enforcement
  • High-throughput Polygon PoS node cluster execution
  • Streams valid JSON-RPC 2.0 response with execution receipts
200 OK Execution ResponseWorkload Execution
HTTP/1.1 200 OK
Content-Type: application/json
X-Satelink-Cost: 30 micro-units
X-Satelink-Latency: 48ms

{
  "jsonrpc": "2.0",
  "id": 1042,
  "result": {
    "number": "0x3f1a24b",
    "hash": "0x9c74b...e0a2",
    "timestamp": "0x68f4e21a"
  }
}
HTTP Status200 OK
Node TargetPolygon Mainnet
07Settlement & Receipt
Step 7/7

Usage Metering & On-Chain Epoch Settlement

Usage is metered at flat rates, settling on-chain to RevenueVault on Polygon PoS.

Every call is logged in the epoch metering ledger. Micro-payments settle cryptographically on-chain to RevenueVaultV2 on Polygon PoS (137) or settle via facilitator on Base, producing an auditable receipt for the agent.

  • Per-call decrement from ledger balance at flat unit rate
  • Periodic batch epoch settlement to RevenueVault contract (137)
  • Verifiable cryptographic receipt issued for accounting
settlement_receipt.jsonSettlement & Receipt
{
  "epoch_id": 8412,
  "account": "0x71C...a4E8",
  "calls_settled": 1,
  "settlement_rate": "flat_unit",
  "vault_address": "0x577D...BaCEF",
  "network": "polygon_pos_137",
  "status": "SETTLED_ON_CHAIN",
  "tx_hash": "0x89e2c...10fa"
}
LedgerEpoch Aggregated
ContractRevenueVaultV2

Control boundary

GO, WAIT or REJECT — verifiable gates for every agent action

Whether executing market analysis or procuring an upstream API, deterministic hard gates verify budget, freshness and authorization before any action occurs.

GO

Every hard gate passed and quotas clear. The proposal or request proceeds through verification.

WAIT

Conditions are not met or evidence is incomplete. No action or spend is initiated.

REJECT

A hard gate failed — stale data, a spending limit exceeded, or an expired mandate. The request fails closed.

Safety boundary

AI proposes. Deterministic risk and your signed mandate decide.

AI proposes

Autonomous models can read context and suggest actions, but have no authority to place orders or exceed limits on their own.

Deterministic risk decides

Every action is checked by fixed rules — kill switches, spending caps, data freshness, and price collars.

Your signed mandate decides

Operations run within explicit parameters authenticated by one-time codes. You can revoke a mandate or pause everything at once.

You keep custody

Your money stays at your broker or wallet. Satelink never holds customer funds, and withdrawals are never enabled.

Risk and how it is limited

Three interfaces

Web for people. API and MCP for AI agents and machines.

Same core, same checks, same receipts — whichever way you connect.

Web

Live

Manage keys, configure spending limits, review transaction receipts, and explore the service catalogue.

API

Live

Programmatic service discovery, metered Polygon JSON-RPC, and Trading Intelligence endpoints over HTTP.

MCP

Not public yet

Model Context Protocol tools for AI agents to discover resources, evaluate data, and propose actions.

Active services

Infrastructure running today on Satelink

The payment, data and execution infrastructure Satelink runs today, for developers and autonomous agents.

Machine Commerce

Discover a priced API, pay for it without a human, and settle on-chain.

Learn more

x402 Machine Payments

HTTP-native, keyless pay-per-call payments in USDC on Base.

Learn more

RPC Infrastructure

Metered Polygon JSON-RPC at a flat per-call rate.

Learn more
Questions & Answers

Frequently asked questions

Explore all 20+ questions across all categories

Satelink is programmatic infrastructure for AI agents and automated systems to discover digital services, evaluate options, and access APIs through controlled, machine-readable workflows. It is operated by Jakuraa Commercial Private Limited.

Paper trading in controlled beta. Live execution is not yet available.

Build on Satelink machine-commerce infrastructure.

Request beta accessTrading AI overview

Not investment advice

Nothing on these pages is investment advice or a recommendation to buy or sell anything. Trading can lose money, including all of it. Simulated and hypothetical results do not predict real results, and nobody can promise any outcome.

Independence

Binance, Upstox and Alpaca are named only to describe technical compatibility work. Satelink is not affiliated with, endorsed by or a partner of any of them, and no broker, exchange or regulator has reviewed or approved Satelink or Satelink Trading AI.

AI disclosure

The agent uses third-party AI language models. AI output can be wrong, incomplete or out of date, so the agent can only read data and suggest; a person approves every order.