Agentic Clearing House

The fiduciary layer for agentic payments.

Skalor is an agentic clearing house. A real-time control layer for AI agents. Detect fraud. Contain rogue agents. Audit every action and payment.

For humans →For agents →
skalor · risk engine
// rogue agent detectedPOST /agent/charge amount: $48,000.00 vendor: unknown-api.xyz // skalor evaluates · 12 ms→ daily_cap FAIL $48k > $1k→ vendor_allow FAIL not on allowlist→ velocity FAIL anomaly detected403 BLOCK · agent quarantined
01 · Why enterprises need this

Your agents are already spending real money. Your finance team has no controls.

$15T
B2B purchases intermediated
by AI agents by 2028 · Gartner
The gap in your stack
Identitysolved · KYC / DID
Railssolved · x402 / Tempo / Canton
Discoverysolved · MCP / agent dirs
→ Fraud & risk controlmissing
One rogue agent — and your monthly budget is gone before anyone wakes up.
02 · What Skalor is

A real-time control layer for AI agents. Detect fraud. Contain rogue agents. Audit every action.

Step 01
AI agent
Initiates an autonomous action — payment, transfer, claim, decision
→
Step 02 · Skalor
Real-time risk check
Fraud detection · policy · anomaly · containment
→
Step 03
Target system
Internal APIs, payment rails, external services
→
Step 04
Outcome
Allow · block · escalate — with full audit trail
Hard limits
Per-agent, per-vendor, per-day, per-action caps. Stop financial exposure before it starts.
Allowlists
Pin agents to approved vendors, systems, and data sources. Default-deny for everything else.
Anomaly & circuit breakers
Auto-quarantine on velocity spikes, behavioral drift, or prompt-injection signals. Human review queue.
Audit trail
Every decision signed, hashed, and replayable. SOC 2 ready. Regulator ready.
03 · How it works

One SDK. Drop-in fraud prevention.

Skalor sits in front of any agent action — payments, API calls, transfers, decisions. Define risk policy in code. Every action is evaluated, logged, and either allowed or blocked, in single-digit milliseconds.

Native SDKs for TypeScript, Python, and Rust. Drop-in middleware for Express, Hono, FastAPI. Integrates with x402, internal APIs, and payment rails (Tempo testnet today, Canton coming soon).

<12 ms
risk eval p99
99.99 %
availability target
typescript · agent.ts
import { skalor } from "@skalor/agent";const agent = skalor(({)policy: {dailyCap:   "$1,000",perVendor:  "$200",allowlist: ["api.openrouter.ai", "*.replicate.com"],velocity:  { max: 10, window: "1m" },},rail: "tempo", // testnet today; mainnet & canton soon});// every agent action runs through risk policy firstawait agent.charge(({ vendor, amount }));
04 · Use cases

Built for the enterprises with the most to lose.

CASE / 01
Banks
Customer-service and treasury AI agents handling transfers, refunds, and counterparty payments. Skalor enforces per-agent transaction limits, blocks anomalous flows in real time, and produces the audit trail your fraud team and the regulator will demand.
→ Real-time fraud detection · transaction limits · audit
CASE / 02
Insurers
Claims-processing, underwriting, and customer-service AI. Skalor stops agents from approving fraudulent claims, holds them to policy ceilings, and gives compliance a complete decision audit for every autonomous action.
→ Claim fraud detection · policy enforcement · decision audit
CASE / 03
AI platforms
Companies running agents at scale — internally or for customers. Skalor sits at the platform layer, enforcing tenant-specific policy, containing compromised or prompt-injected agents, and giving trust & safety the controls they need.
→ Multi-tenant policy · prompt-injection containment
CASE / 04
Developers
Engineering teams shipping agent products. Stop rebuilding fraud detection, audit trails, and policy enforcement from scratch. Drop in Skalor as middleware and ship safe agents in days, not quarters — without owning the compliance plumbing.
→ Drop-in SDK · policy as code · TypeScript / Python / Rust
05 · Trust & compliance

Built for the risk and compliance teams that have to sign off on agents.

Agent commerce will be regulated. Agent fraud is already happening. Skalor is built for both: deterministic policy, signed audit trails, and a non-custodial model that risk teams can actually approve.

  • Custody
    Non-custodial. Keys held by the user or partner KMS. Skalor never touches funds.
  • Audit
    SOC 2 Type II in progress. Independent third-party security audits underway.
  • Networks
    Currently on x402 testnet via Tempo. Mainnet and Canton support coming soon.
06 · Developers

Open spec. Open SDKs. Launching with public beta.

Coming Soon
Quickstart
Coming Soon
Documentation
Coming Soon
GitHub
Coming Soon
Discord

Run agents at scale. Sleep at night.

Early access is open to a small cohort of banks, insurers, and AI platforms running agents in production.