Axiom
Permissionless AI
An on-chain AI inference service. Any agent or dApp calls it, pays per-query via x402, Venice runs the AI, and 1Shot settles gas in USDC — no native ETH required, no backend needed.
Per query cost
Avg inference latency
Native gas required
Networks supported
How it works
One query. Four steps. Fully autonomous.
Call
Agent or dApp sends an inference request to the Oracle endpoint
Pay
x402 HTTP payment header triggers a micro-payment via ERC-7710 delegated permission
Infer
Venice AI runs the query — text, image, or multimodal — with permissionless intelligence
Settle
1Shot relayer executes the on-chain transaction, gas paid in USDC via 7702 smart account
Who it's for
Built for builders who move fast.
- Pay per call with no pre-auth
- ERC-7710 delegation for recurring spend
- No API keys to rotate or store
- Works with any agent framework
- Drop-in AI endpoint for your dApp
- Onchain payment — no Stripe, no subscriptions
- USDC settlement on 11 EVM chains
- Open-source, self-hostable
- Reference implementation of x402 + 7710
- Venice AI as a censorship-resistant backend
- 1Shot for gasless UX without ETH
- Composable with any ERC-4337 stack
Tech stack
Built on open standards.
Permissionless text, image, audio, and multimodal AI. Powers all inference inside the Oracle.
HTTP 402 payment-required standard. Agents pay per-query with no pre-approval or subscription.
Smart account delegations with caveats. Agents get scoped spend permissions — no private keys shared.
Fine-grained permission requests. Users approve exact spend limits, time windows, and target contracts.
Executes 7710 delegated transactions on mainnet with gas paid in USDC, USDT, or USDG.
EIP-7702 upgrades regular EOAs to smart accounts. Enables delegation and advanced permission flows.
Start inferring in 60 seconds
Connect your MetaMask, fund with 1 USDC on Base, and run your first permissionless AI query.