Skip to content
#

spending-controls

Here are 14 public repositories matching this topic...

Spending guardrails for AI agents: the LLM proposes, a deterministic policy engine decides. Budgets, caps, human approvals and an audit trail, paid through the PayPal sandbox. Includes an attack benchmark vs naive and prompt-hardened agents.

  • Updated Nov 16, 2026
  • TypeScript

🛡️ AgentTill: the trust layer for agentic commerce. An autonomous buyer agent that plans purchases, attempts checkout, and enforces deterministic spending policies while recording every decision in an append-only, Merkle-verified ledger.

  • Updated Sep 5, 2026
  • JavaScript

Draft v0.1 vendor-neutral JSON Schema for an AI agent's spending policy (caps, allowlists, approval thresholds), with a field-by-field mapping to 14 payment providers' native settings. Not a standard; not endorsed by any provider. CC BY 4.0 docs / Apache-2.0 code. By Pink Agentic AI Payments.

  • Updated Oct 7, 2026
  • JavaScript

Leash — Spend controls & audit trail for AI agents that pay with PayPal. A control layer between agents and PayPal: every payment request is evaluated against a human-owned policy (budget, allow-list, limits) returning APPROVE, DENY, or ESCALATE. All decisions written to an append-only ledger. Sandbox prototype for PayPal AI Hackathon 2026.

  • Updated Nov 13, 2026
  • Java

Allowance — independent third-party profile of a public API surface, by API Evangelist. Allowance is a consumer trust layer for AI agent payments, backed by Y Combinator. It lets people give AI agents limited, revocable spending power without ever sharing a real credit card number.

  • Updated Oct 28, 2026

Add this topic to your repo

To associate your repository with the spending-controls topic, visit your repo's landing page and select "manage topics."

Learn more