Economic control for AI agents.

Valto is pre-execution policy and a decision ledger for AI-initiated economic actions. Before an agent spends money, Valto helps decide whether it should.

Get started View on PyPI

Why Valto?

Payment rails answer: can this transaction happen? Valto answers: should this agent take this economic action right now?

Agents are beginning to buy data, call paid APIs, trigger compute, and make x402 payments. They need budgets, policies, risk checks, and outcome learning before money moves.

Install

pip install "valto[server]"

Quickstart

from valto import Valto

valto = Valto()

decision = valto.decide(
    agent_id="research-agent",
    action="paid_api_call",
    amount_usd=0.25,
    reason="Buy market data for current task"
)

if decision.approved:
    # execute the economic action
    result = "purchased"

    valto.report_outcome(
        decision_id=decision.id,
        outcome="useful",
        value="data improved final answer"
    )

x402 demo

The first benchmark shows an agent spending money using x402 payments on Base Sepolia, with and without Valto in the loop.

Without Valto Agent directly pays for tools and data.

With Valto Agent requests approval before each economic action.

Outcome ledger Each spend is later labeled useful, wasteful, duplicate, or risky.

Benchmark

Coming soon: repeated agent simulations measuring spend quality, duplicate purchases, budget violations, useful spend ratio, and final task/business outcome.

Product direction

Valto starts as an SDK and local policy layer. The larger product is a managed control plane for agent economic actions: policy, approvals, budgets, risk scoring, decision history, and outcome-driven tuning.