Protocol Live on Hedera Testnet

AgentRouter

A spot market where AI agents buy LLM inference per request in HBAR on Hedera. Low latency. Zero trust. Infinite scale.

databaseTotal Volume
boltRequests Served
hubProviders Live
paymentsAvg Price

Protocol Architecture

01
dns

Providers list

  • check_circleRun the provider service on your local machine or any cloud box.
  • check_circleStake 50 ℏ to the escrow account as a quality bond.
  • check_circleRegistration is automatically posted to the HCS registry topic.
02
shopping_cart

Agents buy

  • check_circleAgent requests a specific model from the global exchange.
  • check_circleExchange routes the request to the cheapest available provider.
  • check_circlex402 protocol fires: pay ℏ → receive LLM completion.
03
gavel

Verifier enforces

  • check_circleRandom sampling: prompts are replayed against a shadow provider.
  • check_circleDivergence in output tokens triggers a verification event.
  • check_circleEscrow stake is seized and the cheater is dropped from routing.

Monetize Your Compute.

Turn your idle RTX or H100 into a high-yield asset. Join the decentralized backbone of AI inference and earn HBAR for every request served.

memory
Minimum Hardware
Any box that can reach the Groq API — no local GPU required
verified_user
Quality Bond
50 ℏ stake — slashed if you serve a cheaper model than advertised
provision.sh — bash
# 1. Clone and install
git clone github.com/GigaHierz/ETH-global-lisbon-2026 && pnpm install
# 2. Add your Hedera keys to .env
HEDERA_PROVIDER1_ID=0.0.xxxxx · HEDERA_PROVIDER1_KEY=0x…
# 3. Make your box reachable
PROVIDER_PUBLIC_URL=https://your-tunnel-or-vps
# 4. Start — stakes 50 ℏ + registers on HCS automatically
pnpm provider1
// unpaid request → the paywall answers
curl -X POST localhost:4021/v1/chat/completions \
-d '{"model":"llama-3.3-70b-versatile",…}'
HTTP/1.1 402 Payment Required
scheme: exact · network: hedera:testnet
amount: 10000000 tinybars · payTo: 0.0.9744152