Halo Halo
English
Use Halo Now
← All guides

Consume inference with your agent

Let an AI agent set up Halo by chatting — pick a model, choose confidential mode, fund a wallet, and it has pay-per-prompt inference with no API keys.

AgentsGetting started

You don’t need to touch a terminal. If you have a capable AI agent — OpenClaw, Claude Code, Hermes, or any agent that can read a URL and run commands — it can set up Halo for you through a normal conversation, and then use Halo to pay for its own inference in USDC on Base. No accounts, no provider API keys.

Halo is in alpha on Base mainnet with real USDC.

Start by pointing your agent at Halo

Paste this to your agent:

Read https://app.runhalo.xyz/skill.md and follow the instructions

From there it’s a conversation. Your agent will ask you a few things — or you can tell it up front:

  • Which model to use. Browse what’s available on the network at app.runhalo.xyz/models and tell your agent which model you want it to use.
  • Whether you want privacy. If your prompts are sensitive, tell your agent to use a confidential model (TEE) — the request is encrypted to a hardware enclave so even the operator serving it can’t read your prompt.
  • Spending limits. Tell your agent a per-prompt and/or overall USDC budget so it stays within what you’re comfortable spending.

Fund the wallet

Your agent creates its own wallet. Ask your agent for its wallet address, then send it:

  • USDC on Base — this is what pays for inference.
  • A small amount of ETH on Base — used only once, for provisioning your Halo balance (the vault deposit). After that, gas is sponsored by Halo, so your agent never pays a network fee while prompting.

Tell your agent once the wallet is funded, and it will provision the balance and start using Halo.

That’s it

Your agent now has a paid, pay-per-prompt inference endpoint it can call on its own — using the model you chose, private if you asked for it, capped at your budget. No keys to manage, no provider signups.