x402 · MCP · USDC on Base

Safety tools for AI agent payments

PayStage sits in front of an agent's wallet and answers one question before money moves: should this payment happen? Preflight checks return pay, review, or do_not_pay.

Early stage. Packages are Available soon — not yet on npm.

Tools

Small, focused pieces that compose into a payment safety layer.

Policy engine

Roadmap

Declarative rules for spend limits, allowed payees, and allowed networks. One policy.json per agent.

Payee allowlist

Roadmap

Known-good payees pass. Unknown payees route to review instead of silently paying.

Spend guardrails

Roadmap

Per-call and rolling budget limits so a runaway agent can't drain a wallet.

Audit log

Roadmap

Every verdict recorded with the inputs that produced it, for later review.

Roadmap

  1. Available soon

    Preflight MCP server

    Policy-driven verdicts for x402 payments over MCP. Not yet on npm.

  2. Roadmap

    Policy engine

    Richer rules: per-asset limits, time windows, per-tool budgets.

  3. Roadmap

    Payee reputation

    Signals beyond the allowlist for deciding when to escalate to review.

  4. Roadmap

    Audit and reporting

    Structured logs and summaries of what an agent paid and why.