Agents

Give your agent
a budget.
Keep control.

An SDK so your agents execute payments in production inside a financial guardrail: a maximum budget, rules they can't skip, a kill switch at any moment.

Talk to the founders →30 min. Founders and technical team.

Why this

Building the agent is the easy part.
Deploying it with real money
is what no one solves.

If you're building agents that make spending decisions (procurement, payment ops, customer refunds, internal automation), the question isn't whether the agent works. It's:

Q · 01How much can it spend before it needs human approval?
Q · 02If it gets it wrong, who stops it, and how fast?
Q · 03How do you audit what it decided, when, and under what authorization?
Q · 04How does it integrate with the financial controls your company already has?

Without an answer to those questions, deploying an agent with real money is a decision no finance team signs off on. That's why most AI pilots stay in the sandbox.

What the SDK provides

The financial guardrail,
in code.

01 / Programmable budget

Programmable budget.

Define how much the agent can spend: a max per operation, aggregate spend per period, spend by category. We evaluate it before every move. If it exceeds, it doesn't execute.

02 / Immutable rules

Rules the agent can't skip.

Whitelist of authorized counterparties. Allowed operation types. Time windows. Rules apply before execution, not after. The agent can't overwrite them.

03 / Kill switch

Kill switch.

Stop any agent in one click from the dashboard. No code. No waiting for an engineer. The operation is cut before the next move.

04 / Audit log

Full audit log.

Every decision is recorded: what the agent did, how much it spent, under which rule, against which event triggered it. Exportable, immutable, signable.

How it's used

Three calls.
An agent with money.

procurement-bot.ts

TypeScript · v1

// 1. Import the SDK
import { Relayer } from "@relayer/sdk";

// 2. Create an agent with its guardrails
const agent = relayer.agents.create({
  name: "procurement-bot",
  budget: { daily: 5000, currency: "USD" },
  allowlist: ["supplier-a", "supplier-b"],
});

// 3. The agent declares a payment intent
await agent.pay({
  to: "supplier-a",
  amount: 1200,
  currency: "USD",
  reason: "monthly subscription",
});

The SDK handles applying the rules, validating the budget, executing the payment via the right corridor, and recording the audit log. Your agent just declares the intent.

Use cases

Where we're already seeing demand.

· Procurement

Procurement automation.

Agents that execute payments to recurring suppliers within budget, escalating to a human when there are exceptions.

· Customer ops

Customer operations.

Agents that process refunds, adjustments, and credits within rules, without waiting for human approval on routine cases.

· Internal

Internal automation.

Agents that move funds between accounts of the same company based on operational conditions: rebalancing, top-ups, sweeps.

Current status

Where we are today.

The SDK is ready and running in test environments. We're looking for the first technical teams to build production pilots: companies with clear use cases, aligned finance teams, and a willingness to iterate the operating framework with us.

If your team is deploying agents with real money and needs the financial control layer before going to production, let's talk. The first conversation is with the founders and our technical team.

Founders + technical team

Thirty minutes with the founders and the technical team.

For your technical team, a conversation about architecture, integration, and the SDK. For your finance or compliance team, a conversation about controls, audit, and the operating framework. We do it separately or together.