run.pay started as a way to pay for API calls. It grew into real financial tooling for any business running more than one autonomous agent — spending controls, risk scoring, multi-rail payment routing, and complete audit trails.
This is tooling layered on top of the same marketplace you already know — not a separate product, not a new account to create. Everything here is optional and additive.
Set real limits before an agent ever makes a call — never after the fact.
A unified check before every payment — trust score, spending limits, and approval thresholds combined into one clear signal: proceed, review, or block.
POST /api/guard/evaluate
Not just "can this agent spend this" — "is this actually what it said it was going to do." A real check against context manipulation and prompt injection in agent payment flows.
POST /api/intents/declare
A human cryptographically pre-authorizes exactly what an agent may spend — amount, category, minimum vendor trust score — before any call happens.
POST /api/agents/mandates
Freeze a specific agent, a specific vendor, or the entire platform instantly. The button you hope to never need.
POST /api/admin/kill-switch/freeze-all
Real visibility across every agent you run, without building your own dashboard.
Identity, wallet status, reputation, active mandate, and recent activity for any agent — in one view. Plus a fully explainable trust score, never an opaque number.
GET /api/agents/:id/passport
Group agents into cost centers — marketing, research, procurement — and see what each one has spent and has left, without building it yourself.
GET /api/agents/:id/treasury
Spending broken down by category, formatted receipts per transaction, and a single view of balance, today's spend, failure rate, and pending approvals.
GET /api/agents/:id/dashboard
A complete, tamper-evident chain from declared intent to vendor selected to final payment — the real "why did this agent spend this" answer.
GET /api/observability/:tx_id
One integration, multiple settlement rails — chosen automatically.
Card, stablecoin, or native crypto — run.pay recommends the right rail for the amount, never forcing one path.
POST /api/pay/route
Direct on-chain USDC settlement, verified and settled by run.pay itself — no third-party facilitator.
A real settlement rail, tested end-to-end on testnet — never the headline, just one more honest option. Mainnet activation pending, not yet available for real funds.