Paitify — Agent Spend Policy Engine
Loading…
Real-time authorizationYour AI agents spend with guardrails.<br>Paitify is the spend policy engine for autonomous AI agents. Define granular rules, authorize transactions in real time, and maintain a complete audit trail — without slowing your agents down.<br>Get started freeSee the API
auto_approve authorize_spendCaptured<br>"state": "CAPTURED",<br>"authorizationId": "b2f1e4a0-…",<br>"amount": 249.99,<br>"merchant": "AWS"
auto_approve Amount (USD)$
Merchant<br>AuthorizeEnter an amount and hit Authorize to see a live decision.
REAL-TIME DECISIONS·99.9% TARGET UPTIME·UNLIMITED POLICIES
See it in action<br>90 seconds: watch an AI agent get auto-approved at $49 — and stopped at $600.
AI agents can spend money. A lot of money.<br>Modern LLM agents browse, purchase, spin up infrastructure, and call paid APIs — often autonomously, 24/7. Without centralized spend controls, a single misconfigured agent can drain budgets, violate compliance policies, or transact with unauthorized merchants before anyone notices.<br>Traditional payment controls were designed for humans. Paitify is built for machines — real-time decisions, policy as code, and machine-readable JWT tokens your payment layer can verify without calling home.
Runaway spend<br>Agents loop and retry, spending 10× more than budgeted before anyone sees an alert.
No policy enforcement<br>Hard-coded limits in agent prompts are fragile and can't be updated without redeploys.
Zero audit trail<br>Finance teams have no visibility into which agent spent what, where, and why.
Use cases<br>Whatever your agent buys, and whatever rail it pays on, Paitify is the policy and audit layer that decides yes or no — and logs it.
Procurement agents<br>An agent buying software, supplies, or cloud resources against a department budget.<br>$2,400 order → REQUIRES_APPROVAL
API & cloud cost control<br>A coding agent calling metered APIs, LLM tokens, or compute against a hard ceiling.<br>$40 API bill → CAPTURED
Ad-spend & marketing agents<br>An agent adjusting bids or launching campaigns within daily and campaign budget caps.<br>$150 daily spend → CAPTURED
Trading agents<br>An agent executing trades within position and value limits, logged for reconstruction.<br>$5,000 trade → DENIED
See all 13 use cases
How it works<br>One API call before every payment. Policies enforced. Every decision logged.
01<br>Define your policies<br>Set spend limits, merchant rules, and approval thresholds in the dashboard — per agent or company-wide.
02<br>Agent calls /authorize<br>Your agent makes one API call before each payment. Paitify approves or denies in real time, with a reason.
03<br>Review the audit trail<br>Every decision is logged automatically. Review the full audit trail in your dashboard.
authorize_spend.py<br>import requests
response = requests.post(<br>"https://api.paitify.io/v1/integration/authorize",<br>headers={"X-API-Key": PAITIFY_API_KEY},<br>json={<br>"agentId": "procurement-agent-01",<br>"amount": 249.99,<br>"currency": "USD",<br>"merchantName": "AWS",<br>"mccCode": "7372",
result = response.json()
# result["state"] is the sole source of truth for the decision.<br>if result["state"] == "AUTHORIZED":<br>charge_card(amount=249.99)<br>confirm_capture(result["authorizationId"]) # POST /v1/integration/authorize/{id}/capture<br>elif result["state"] == "CAPTURED":<br>charge_card(amount=249.99) # AUTO capture mode: already recorded — skip /capture<br>elif result["state"] == "REQUIRES_APPROVAL":<br>wait_for_approval(result["authorizationId"]) # poll GET /v1/integration/authorizations/{id}<br>else:<br>print(f"Spend denied: {result['reasonCode']}")
Built for production AI systems<br>Every feature designed for the real operational challenges of deploying autonomous agents at scale.
Decisions in the request path<br>Redis-backed spend counters and in-memory rule evaluation. No batch delay — every authorization is checked live.
Flexible capture modes<br>AUTO: spend commits immediately on authorize, call /cancel only on failure. EXPLICIT: reserve with AUTHORIZED, commit with /capture, or release with /cancel.
MCP-native<br>Works with Claude and any MCP-compatible client out of the box.
Rich audit trail<br>Every decision logged with full rule evaluation results, agent ID, merchant data, and timestamps.
Human-in-the-loop<br>Flag high-value transactions for human approval. Agents wait; humans decide; budget holds until resolved.
JWT-signed approvals<br>Approvals are RSA-2048 signed tokens. Payment processors verify offline using JWKS — no callback latency.
Granular policies<br>Per-agent or company-wide. Limits by amount, period, MCC code, currency, velocity, merchant, and business hours.
Velocity controls<br>Rate-limit transactions per sliding window. Stop looping agents before they hit your card limits.
Protocol-agnostic<br>REST API. Works with any stack, any language, any payment processor.
Every decision, chained and provable.<br>Every authorization Paitify makes — approved, denied, or sent for human review — is written to an append-only audit log in the same transaction as the decision...