TreasuryPlatformReconciliationPricingFAQTalk to the founders →

FINANCIAL GUARDRAILS FOR ASSET MANAGERS AND OPERATORS

Procurement for AI agents

Spending limits, approval policies, and audit trail for the AI agents in your operations. Per client, per strategy, per workflow.

Talk to founders →Read the docs

SDK or dashboard · Live in two weeks · Self-custodial wallets per agent

LLM TOKENS
$4,200
monthlyOn track
API CALLS
12,400
callsNormal
ON-CHAIN PAYMENTS
$5,800
executed$6.5K budget

WHEN AI ENTERS PORTFOLIO OPS

Set the controls before AI scales across clients
Better designed in than retrofitted later

When AI starts running research, allocation, or reporting across multiple client accounts, attribution, audit, and reconciliation stop being optional. Build the controls before the first agent touches a client account, not after.

01
COST ATTRIBUTION BY CLIENT

AI cost rolls up to the right client

When your team uses AI for research, allocation, or reporting on behalf of a specific client, the cost lands on that client's P&L, not as a fund-level overhead. Tokens, API calls, and on-chain fees auto-attribute to the right account.

02
AUDIT TRAIL COMPLIANCE-GRADE

An audit trail your compliance team can defend

Every AI action (query, allocation, report, transfer) logged with the policy applied, the threshold crossed, and the human who approved. When the SEC, your fund administrator, or your LPs ask, the answer is one export away.

03
RECONCILIATION ON RAILS

Reconciliation without the month-end fire drill

AI activity reconciles automatically against your fund administrator's records, your ERP, and custody statements. What used to be days of CSV matching at quarter-end becomes daily, automatic, exportable.

WHAT EVERY AGENT GETS

An account, a budget, and a trail
Per agent, per client, per strategy

Every AI agent in your operations runs inside a dedicated account you control, with spending limits you set and a full audit trail your compliance team can export. Same setup for every agent, scaled across your client book.

01 — ACCOUNT

A dedicated account per agent that you control

Every agent runs out of its own programmable account, secured by the same institutional infrastructure Coinbase, Circle, and Robinhood use. The agent can spend within your rules, and only within your rules. Funds never leave without policy approval.

02 — BUDGET

Spending limits the agent cannot exceed

Hard caps per agent, per client, per category. Routine spend auto-approves under threshold. Material spend routes to a human approver. The policy is enforced at the moment of action, not at month-end review.

03 — TRAIL

Real-time logs of every agent action

Every query, transfer, or report the agent runs is logged with the policy applied, the threshold reached, and the client it served. Exportable to your fund administrator, ERP, or compliance tool whenever asked.

FOR YOUR COMPLIANCE TEAM

What your compliance team sees
And what you can hand to your fund administrator

Every action your agents take rolls up to a single view your compliance officer or CFO can read in seconds. Per client. Per agent. Per policy. Exportable to your fund administrator or auditor when the question comes.

relayer.fi/complianceEXPORT ▾
MTD AI SPEND BY CLIENT
Client A · Acme Holdings$4,820within
Client B · Vega Capital$3,150within
Client C · Ortelius Family$1,240within
Client D · Three Rivers$2,470approval
RECENT AGENT ACTIONS
Research Agent · Client A · $42 query
→ Auto-approved · Policy 7B · 12:42 PM
Reporting Agent · Client B · $180 generation
→ Auto-approved · Policy 4A · 11:31 AM
Risk Monitor · Client D · $520 transfer flagged
→ Pending CIO approval · 09:14 AM
EXPORT TOFund Admin ↓ERP ↓Compliance ↓

TEMPLATES

Agents your operations could run today
Pre-configured for asset managers

Start with one template that maps to a workflow you already run. Configure it for your portfolio. Scale to more agents as the model becomes routine.

RESEARCH

Due diligence research agent

Pulls data from Apollo, Pitchbook, on-chain sources, and industry databases. Produces investment memos and account briefs for new positions.

CONSUMES
Apollo · Pitchbook · On-chain · Custom DBs
PRODUCES
Investment memos · Account briefs
REPORTING

Monthly client reporting agent

Generates monthly performance reports per client account, with attribution by strategy, fee transparency, and benchmarks.

CONSUMES
Portfolio data · Market feeds
PRODUCES
Per-client PDF reports
RECONCILIATION

Custody and ERP reconciliation agent

Matches transactions across custody statements, fund administrator ledgers, and ERP. Produces a clean reconciled ledger with exceptions flagged.

CONSUMES
Custody · Fund admin · ERP
PRODUCES
Reconciled ledger · Exceptions list
RISK

Position and risk monitor

Watches portfolio positions for drift, drawdowns, and exposure breaches. Alerts the CIO and produces breach reports when thresholds are crossed.

CONSUMES
Positions · Market data · Limits
PRODUCES
Real-time alerts · Breach reports

Build your own? Each template is customizable. Or start blank and define your agent from scratch.

ONE LINE TO INTEGRATE

Wrap your agent. Keep building.
Works with the runtime you already use

Add Relayer.fi to your existing agent stack in one line. Memory, tools, and reasoning stay yours. Relayer.fi adds the account, budget, and audit trail underneath.

agent.ts
import { relayer } from '@relayer/sdk'

// Wrap your agent. Budget, policies, and audit trail apply.
const agent = relayer.wrap(myAgent, {
budget: 5000,
client: 'acme-holdings',
})
RUNTIMES SUPPORTED
MastraLangChainOpenAI SDKAnthropic SDKVercel AI SDKCustom

First account live in two weeks. Onboarding self-serve once you have an API key.

Questions we hear from operators bringing AI to ops

MCP is a model interaction protocol. x402 is an agent payment protocol. Both are layers below Relayer.fi. We add the financial control layer on top: budgets per agent, approval workflows, audit trail, and reconciliation against your accounting stack. We compose with x402 and any MCP, we don't replace them.

Yes, when presented properly. Every Relayer.fi transaction maps to a standard ledger entry with timestamp, counterparty, amount, and policy applied. Major fund admins accept on-chain entries when they're reconciled to fiat-denominated reports, which Relayer.fi generates automatically.

Yes. An agent can pay an API in stablecoin, settle a vendor in MXN via SPEI, and write a record to your ERP, all in a single workflow under one budget. Relayer.fi abstracts the rails so the agent doesn't need to know which one it's using.

No. You can deploy Relayer.fi as the procurement layer first, before your AI agents go live. Set up controls, define policies, then build or deploy agents with the controls already in place. Many operators do this exactly to derisk the deployment.

Your AI ops, governed

Book 30 minutes with the founders. Walk us through how you operate today. We'll show you the dashboard live, and if we fit, the controls are ready before your first agent goes live.

Talk to founders →

No cost. No commitment. If we don't fit, we'll tell you on the first call.