Effective Intelligence

Four surfaces

One question runs through all of them: what is this machine worth right now, net of the power it burns, and is that more than leaving it asleep. Everything below is a different way of asking it or acting on the answer.

01

The agent

Runs on your machine. Reads the hardware rather than trusting its label, prices every destination it could serve in dollars per day, and moves the machine when the gain clears the cost of moving it. Dry-run by default until you hand it the keys.

npx effective-intelligence scan How it decides
02

The index

Most of the market has no price feed at all. Illiquid tokens, points that are not tokens yet, networks nobody indexes. Those are scored in-house on real exchange depth and on expected value, priced by contract address and never by symbol, then graded so a bad one is refused rather than discovered.

Read the live index
03

For agents

An MCP server and an onboarding API with no screen in the path, so another agent can enrol a machine, read its scores and act on them without a human clicking anything. Built for Claude, Hermes, OpenClaw and anything else that speaks the protocol.

The MCP surface
04

Settlement

A share of measured uplift, settled in USDC on Base over x402. The fee is the difference between what the machine earned and what it would have earned asleep, so no uplift means no fee. There is no invoice and nothing to cancel.

How payment works
Emission windows, the scorer's method and the FAQ live in the docs. Back to the book