Guides

Start here

Welcome to AgentRail. Read this once and you'll have a clear sense of what the platform is, who it's for, and what to read next.

What is AgentRail?

AgentRail is the financial layer for autonomous AI agents. It gives every agent its own programmable wallet, a policy-aware way to move money, and real-time visibility into everything it does — all behind one coherent API and one conversational interface.

You can think of AgentRail as the equivalent of an issuing-and-acquiring platform, but designed from scratch for software that acts on its own. Identity, compliance, and audit are part of the foundation, not bolted on.

Who it's for

AgentRail is built for teams shipping agents that need to move money in production. We've focused on a few use cases first, but the primitives generalize:

  • ─Logistics & supply chain — pay carriers, warehouses, and fuel vendors as deliveries complete.
  • ─Home services & property management — settle recurring vendor payments and one-off jobs without human dispatchers.
  • ─Marketing & performance — agents that buy ad inventory, pay influencers, or settle creator payouts.
  • ─Financial services back-office — reconciliation, AP automation, and KYC remediation handled by agents under policy.
Note
If your team is shipping agents into production and is currently writing its own integrations against rail-specific SDKs, AgentRail is for you.

What you'll build

After reading these docs, you'll be able to:

  • ─Provision an agent with a wallet, capabilities, and spend policies.
  • ─Send and receive payments across ACH, RTP, FedNow, wires, and cards.
  • ─Stream what your agent is doing into your own observability stack.
  • ─Use AgentRail Chat to operate your agents without a dashboard tour.

The shortest possible “hello world” — using the Python SDK — looks like this:

hello.pyPython
import os, agentrail

client = agentrail.AsyncClient(api_key=os.environ["AGENTRAIL_KEY"])

async def main():
    agent  = await client.agents.create(name="hello-bot")
    wallet = await client.wallets.create(agent_id=agent.id)
    pay    = await client.payments.create(
        wallet_id=wallet.id,
        recipient="acme-logistics",
        amount=4250_00,
    )
    print(pay.status, pay.id)

See Quickstart for the full walkthrough including sandbox setup, idempotency, and webhook verification.

How we think about it

Three commitments shape every product decision. Skim them — they explain why AgentRail looks the way it does.

Agents are economic actors

Every agent gets its own persistent identity, its own wallet, and its own policy envelope. You can audit, pause, or revoke any one of them without touching the rest.

Money is a programmable surface

Rails — ACH, RTP, FedNow, wires, cards — are implementation details. The API talks in payments. AgentRail chooses the right rail under the hood, and tells you which one it picked.

Compliance is part of the platform

KYC, KYB, sanctions, and audit are paid for by per-transaction pricing. You ship faster, regulators get what they need, and you never have to plumb a third-party screening tool.

Where to go next

  • ─Quickstart — five minutes to a real payment in sandbox.
  • ─Core concepts — wallets, agents, policies, and how they fit together.
  • ─Using AgentRail Chat — prompt patterns that work, and the ones that don't.
  • ─Webhooks — event payloads, retry semantics, signature verification.
← Back to docs indexAsk AgentRail Chat →