Skip to main content

Serve AI on the open market.
No permission needed.

Set your price. Announce to the network. Receive USDC for settled deliveries, depending on demand, availability, successful settlement, and your configuration — whether you run a model, a routing service, or a specialized agent.

Three ways to provide

All three serve buyers on the open market. What runs behind is entirely yours.

Raw Inference

You run a model or proxy an upstream API — Ollama, a fine-tune, a local GPU, OpenAI, Together. Point AntSeed at it with one config entry and announce it to the network. Buyers choose you based on price, latency, and on-chain reputation; payments depend on demand and successful settlement.

  • → Any model or backend
  • → Set your own price per token
  • → Reputation built per delivery
Become a provider →

AI Agent

You've built domain expertise in AI form. A legal agent, a security researcher, a trading analyst. Announce it as a named service. Buyers pay for your expertise, not just the tokens.

  • → Persona, guardrails, and knowledge stay private
  • → Announced as a named service on the network
  • → Premium pricing for specialized delivery
Become a provider →

Routing Service

Build specialized routing logic and offer it on the network. Latency-optimized, cost-minimizing, TEE-only, or domain-aware. Receive payment for settled routed requests without running a single model.

  • → No model infrastructure required
  • → Latency, cost, TEE, or domain-aware routing
  • → Payment per settled routed request
Become a provider →

Provider Compliance

AntSeed is designed for providers who build differentiated services — such as TEE-secured inference, domain-specific skills or agents, fine-tuned models, or managed product experiences. Simply reselling raw API access or subscription credentials is not the intended use and may violate your upstream provider's terms of service.

Providers are independent operators and are solely responsible for their models, infrastructure, outputs, logs, privacy practices, data handling, security, sanctions/export compliance, tax obligations, applicable AI laws, and upstream API provider terms.

Seller-side ANTS emissions are currently tracked but locked in a dedicated Provider Pool while AntSeed develops stronger validation and proof systems. Fake usage, sybil behavior, or incentive extraction may be excluded or subject to future slashing.

What the network sees. What stays private.

Buyers see enough to route and verify. Everything else stays on your machine.

Public to the network
Your service names
Your price per token or per request
Your capability tags (TEE, domain, model family…)
Your on-chain reputation score
Your latency percentiles
Your uptime window
Under your control — secure your node
🔒Your backend URL or model provider
🔒Your routing logic and selection criteria
🔒Your system prompt and guardrails
🔒Your RAG sources and knowledge base
🔒Your prompt engineering
🔒Your fine-tune weights

One JSON file. Full control.

No code to write. Point AntSeed at an OpenAI-compatible endpoint, set your prices and categories, and offer capacity to buyers. Become a provider →

1. Configure with the CLI
terminal
# Point at any OpenAI-compatible endpoint
antseed config seller add-provider together \
  --plugin openai \
  --base-url https://api.together.ai

# Announce a service with your price + categories
# (--cached is optional — set it to charge less for
# cached-input tokens when your upstream supports them)
antseed config seller add-service together deepseek-v3.1 \
  --upstream "deepseek-ai/DeepSeek-V3.1" \
  --input 0.6 --cached 0.06 --output 1.7 \
  --categories chat,math,coding

# Start serving
export OPENAI_API_KEY=<your-key>
antseed seller start

Compatible with any OpenAI-API endpoint — your own Ollama, vLLM, a fine-tune, or an upstream you have the right to resell. You are responsible for complying with your upstream's terms of service.

2. Or edit config.json directly
~/.antseed/config.json
{
  "seller": {
    "providers": {
      "together": {
        "plugin": "openai",
        "baseUrl": "https://api.together.ai",
        "services": {
          "deepseek-v3.1": {
            "upstreamModel": "deepseek-ai/DeepSeek-V3.1",
            "pricing": {
              "inputUsdPerMillion": 0.6,
              "cachedInputUsdPerMillion": 0.06,
              "outputUsdPerMillion": 1.7
            },
            "categories": ["chat", "coding", "math"]
          }
        }
      }
    }
  }
}

Your backend URL, API key, and routing logic are intended to remain under your control. The network only sees the service name, price, and categories; you are responsible for securing your node and credentials. Full config reference →

Direct settlement. No invoicing.

Buyers lock funds before a session. You deliver. Settlement executes on-chain automatically. Payment protocol details →

01

Buyer locks funds

USDC is locked in the AntSeedEscrow smart contract on Base before the session starts. Requests flow freely while funds are escrowed.

02

You deliver, receipts are signed

Each request generates a provider-signed receipt with exact token counts, cost, and a cryptographic signature. Both sides have proof.

03

Settlement executes on-chain

On session end (or 10 min idle), the escrow contract computes final cost from signed receipts, sends your payout to your wallet, and refunds unused funds to the buyer.

Your priceYou set it — per input token + per output token
Protocol fee4% — may be directed to ecosystem mechanisms such as reserves, grants, incentives, buy-and-burn, or other community-approved uses
Your payout96% of what buyers pay, direct to your wallet in USDC
Payment methodsBuyers pay in USDC or by card — your payout is always USDC
Settlement chainBase mainnet
wallet management
antseed seller status         # earnings, peers, wallet address
antseed seller stake <amt>    # stake USDC to become discoverable
antseed seller unstake        # withdraw your stake

Your EVM wallet is derived automatically from your node's secp256k1 identity key. Payouts land in it on every settlement — no claim step, no separate wallet setup. Full payment docs →

Build reputation that compounds.

Every delivery is recorded on-chain. Your reputation belongs to your wallet. No platform can revoke it.

Success rate

Percentage of requests delivered and settled on-chain

Latency p50 / p99

Measured per delivery, visible to buyers pre-route

Token accuracy

Signed receipts verify exact token counts on both sides

Uptime

Historical availability across announced service windows

Any buyer can build their own access and routing rules on top of on-chain stats. Providers with strong track records may command higher prices and receive more traffic, depending on buyer and router preferences.

Common questions

Does the network see my backend, model choice, or routing logic?+

The network only sees what you announce: your service names, pricing, capability tags, and on-chain reputation. Your backend URL, model provider, routing strategy, system prompt, and fine-tune weights are intended to remain under your control. You are responsible for securing your own node, credentials, logs, and provider infrastructure.

What provider types can I run?+

Three: Raw Inference (serve a model or proxy an existing API), Routing Service (select providers on behalf of buyers and receive payment per routed request), or AI Agent (wrap domain expertise as a named always-on service). A single node can run all three simultaneously at different price tiers.

Does my node need to run 24/7?+

No. Providers announce uptime windows in their metadata. When you go offline, the network routes around you. Your on-chain reputation persists across sessions.

How do payments actually reach me?+

Buyers lock USDC in on-chain escrow on Base before a session starts. Requests flow freely during the session. When the session ends (or idles for 10 minutes), settlement executes on-chain and USDC lands in your wallet automatically. No invoicing, no billing cycles.

Are seller ANTS incentives claimable now?+

Seller ANTS emissions are currently tracked but routed into a dedicated Provider Pool and locked. They are not freely claimable yet. Future claimability is expected after stronger validation, audit, attestation, and proof systems are introduced, and may be subject to verification or slashing.

Can I use any model underneath?+

Yes. You can wrap Anthropic, OpenAI, Together, Ollama, a fine-tuned model, or any standard API. The network only sees what you deliver — not your backend.

Can I serve multiple capability types from one node?+

Yes. A single AntSeed node can advertise multiple services — raw inference on one model, a routing service with custom logic, and an AI Agent, all at different price tiers. Each service is announced independently to the DHT.

Ready to provide?

Install AntSeed, configure your provider, and offer AI capacity as an independent operator.