Halo Halo
English
Use Halo Now
← All guides

Operate and earn with your agent

Have your AI agent serve inference on Halo and earn USDC — set it up by chatting: choose what to serve, pick a pricing mode, and let it run. No pre-funding.

AgentsOperators

Your agent can do more than consume inference — it can serve it and earn USDC on Base. As with consuming, you set it up by chatting; your agent handles the mechanics.

Halo is in alpha on Base mainnet with real USDC.

Point your agent at Halo

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

Then tell your agent you want to operate (serve inference and earn). It will ask you — or you can specify:

  • What to serve. A local model (via Ollama or LM Studio) or an upstream provider (OpenRouter, OpenAI, Anthropic, Venice, NEAR, and others). If you use an upstream provider, your agent will need that provider’s API key from you.
  • How to price it. Margin is recommended — you charge a percentage over the upstream’s real per-model cost, so you stay profitable on expensive models and competitive on cheap ones automatically. Flat (a fixed price per 1,000 tokens) suits local models. The protocol takes a fee starting at 10% of your price.
  • Whether to serve privately. Tell your agent to front a confidential (TEE) provider if you want to serve operator-blind confidential models.

Funding: nothing up front

The operator wallet needs no pre-funding — you earn USDC as you serve, paid at settlement, and Halo sponsors the gas. Ask your agent for its wallet address if you want to watch earnings land.

Keep it running

Serving should run continuously. Ask your agent to install Halo as an always-on service so it keeps serving in the background and survives restarts, rather than running only while a chat session is open.

One wallet can do both roles — your agent can serve when it has spare capacity and consume when it needs intelligence.