SpendGuard
On-chain spending governance for autonomous AI agents on Stellar.
Spending-policy layer for autonomous AI agents: defines how much, with whom and under what conditions the agent can transact, with everything recorded on a public network.
Overview
SpendGuard works as a smart lock between an AI agent and the digital vault holding the owner's funds. The owner defines the rules (daily limit, max per transaction, approved recipients) and the agent can only operate within them. The rules live on the Stellar network, out of reach of any intermediary server.
The problem
Autonomous AI agents already pay for services in real time through the x402 protocol. The only safeguard against a compromised or tricked agent typically lives on external servers that can be bypassed. There was no standard way to guarantee the agent respects the owner's limits in a public, auditable environment.
Approach
Enables autonomous agents to pay for APIs, data and services safely. Sets an auto-renewing daily spending limit, enforces a per-transaction cap to block expensive calls, only accepts payments to pre-approved recipients, gives the owner an emergency switch to pause the agent at any time, and keeps the balance inside the contract itself, never exposing the agent key.
Impact
Contract deployed on Stellar's public test network, with a test suite covering limits, exceptions and attack scenarios. The demo runs real transactions with cap rejection, daily-limit exhaustion and emergency-switch activation. Submitted to Stellar Hacks 2026, AI and Robotics track.