▲ $0.005 per call · x402 USDC on Base L2 · 80% to GPU seeders

BitBrain

BitTorrent for LLMs. Huge models need huge compute — now anyone can seed their GPU into the swarm, and AI agents pay per call. No accounts, no inbound ports, every payment settled on-chain.

1 · SeedRun one command on any machine with a GPU. Your node pulls work over plain HTTPS — home NAT, dorm room, whatever.
2 · ServeYour model answers agent requests. You keep 80% of every call, paid to your wallet automatically on Base.
3 · VerifiedChallenge probes prove you run the model you listed. Honest seeders build trust; fakes get routed away.

Live swarm

Auto-refreshes every 10s.

ModelSeederTrustCallsAvg latency

Pay for a completion (agents)

curl -X POST https://x402-agent-pay.com/bitbrain/infer   -H 'Content-Type: application/json'   -H "X-Payment: $PAYMENT"   -d '{"model":"qwen3-14b","messages":[{"role":"user","content":"Tell me about BitTorrent."}]}'

Missing or invalid payment returns HTTP 402 naming the facilitator. No online seeder for your model returns 503 and charges nothing. A call that no seeder completes is auto-refunded on-chain. Every response carries a proof-of-inference.

Seed your GPU

curl -fsSL https://x402-agent-pay.com/bitbrain/install.sh | bash -s --   --token YOUR_SELLER_TOKEN --model qwen3:1.7b

Registration is curated at launch — email [email protected] with your model, GPU, and payout wallet to get a seller token. The node is a single stdlib-only Python file: ollama in, HTTPS out, nothing else.

Learn more

Model catalog · Seeder registry · Public ledger · OpenAPI · MCP manifest