Agent payments · identity · privacy — all local

Your agent is about to spend your money.
Give it a budget it can't blow, a name it can prove, and a filter on what it shares.

A small proxy that runs on your own computer, between your AI agents and the internet. It pays automatically with stablecoin under the limit you set — free and open source. Add-ons let it cryptographically prove which agent made a payment, and automatically catch sensitive data before it ever leaves your machine.

$ brew install aor && aor init
~/.aor created · Apache-2.0 · runs on your machine
 
$ aor agents create
→ wallet, proxy port, daily limit, per-call max — all yours to set
→ written ~/.aor/agents/pricing-bot.yaml
 
$ aor start && export HTTP_PROXY=http://localhost:8402
# your agent just makes normal requests — no SDK, no code change
 
$ curl https://api.flightdata.io/quote
→ payment required → paid automatically → under budget → cleared ($0.02)
 
$ aor-pro identity setup --agent pricing-bot # add-on
→ generates a cryptographic identity, signs every payment as this agent
 
$ aor-pro cargo setup --agent pricing-bot # add-on, catches sensitive data
→ emails, card numbers, and API keys are now blocked or redacted automatically

Every other option assumes something you don't have

A compliance team. A bank partnership. Or time to build all of this yourself.

CARD NETWORKS / TRADITIONAL PAYMENT RAILS

Built for a human at the switch

Minimums and manual approval hold up for one payment. They don't clear a $0.02 charge a thousand times an hour, and they weren't built to check who's actually asking.

ENTERPRISE AGENT PLATFORMS

Built for a company, not a desk

Identity checks, spend policy, and data-loss prevention are the right size once you're big. At two people, it's a sales call before your first payment clears.

DIY WITH RAW APIS

No guardrails unless you build them

Wire a payment SDK to your agent yourself, and your first spend limit — and your first accidental data leak — is whichever bug you haven't hit yet.

One proxy your agent already knows how to talk to

Point your agent's HTTP client at it. No SDK, no rewritten code.

01

Install it

Runs quietly on your own computer in the background. No account needed to start, and we never hold your money or see your traffic.

brew install aor
02

Set a budget, start paying

A spending cap is on by default the moment you create an agent. Stablecoin payments under the limit clear automatically. Nothing over it does — free, open source, no account required.

daily limit: $5 · per-call: $0.10
03

Prove it, protect it

Two add-ons plug into the same proxy: one signs every payment so you can prove which agent made it; the other catches emails, card numbers, and API keys before they leave the machine. · paid add-ons

aor-pro identity setup && aor-pro cargo setup

Three things, one proxy

Each one plugs into the same request path. Turn on only what you need.

FREE · OPEN SOURCE

Payments

Your agent's requests get paid automatically with stablecoin whenever a site asks for payment — never over the budget you set, per agent.

  • Daily / weekly / monthly limits, plus a hard per-call cap
  • Rate limits, so a bug can't fire off hundreds of payments a minute
  • A local log of every payment, queryable from the command line
  • Your wallet key stays encrypted on your own disk
PAID ADD-ON

Identity

Every agent gets its own cryptographic identity. Payments are signed with it, so you can prove which specific agent made which payment — and shut one off instantly if it's ever compromised.

  • Only agents you've explicitly trusted can use a given wallet
  • Revoke a compromised agent's access in one command
  • Payments carry proof of origin — useful for your own records today, and built on an open, published format so others can check it too, in time
PAID ADD-ON

Privacy

Outbound requests are checked for sensitive data before they're sent anywhere — emails, phone numbers, card numbers, and common API-key formats are caught automatically.

  • Choose per rule: block it, redact it, or just get notified
  • Runs on your machine — the content never has to leave to be checked
  • Add your own custom rules for anything specific to your use case

Every agent ships with a sensible default. Rewrite it per agent.

Most people never open the settings file below. It's there if you want it.

DEFAULT

What you get right away

Limit per payment$50.00
Ask me first above $200Human approval for anything over this
Allow brand-new recipientsOff means only sites you've approved before can get paid
POWER USER

The config file

A plain YAML file per agent. This one has all three pieces turned on, stacked so identity is checked first, then privacy, then the payment itself goes out.

rails: identity_gate: # checks who's calling, first wraps: cargo_gate peers_file: ./trusted-peers.yaml wrapped_config: # privacy rules run next, before anything is sent rules: - {type: email, action: redact} - {type: credit_card, action: block} wraps: x402 # then pays — free, open source wrapped_config: wallet_address: 0xYourWallet daily_limit_usd: "5.00" per_call_max_usd: "0.10"

One dashboard, every agent

A local web page — not a cloud service — showing spend, who signed what, and what got blocked.

AgentOnRails / Dashboard
● running
AgentTodaySigned byStatus
pricing-bot $0.86 / $5.00 did:key:z6Mk…7Fd allowed
pricing-bot did:key:z6Mk…7Fd redacted · email
research-agent $0.14 / $2.00 did:key:z6Mv…2Aq allowed
research-agent unsigned blocked · no identity

The proxy runs on your computer. Not ours.

Local-first by design — self-hosted, free at the core. That doesn't change as it grows.

Runs locally

It lives on your own machine as a background process. Nothing routes through our servers by default.

We never touch your money

Payments move directly between your agent and whoever it's paying — signed and forwarded, never custodied by us.

Your keys stay yours

Wallet and identity keys are encrypted at rest with AES-256-GCM + scrypt, on your disk — never sent anywhere to unlock.

Open source at the core

The proxy and the free payment engine are Apache-2.0. Every payment lands in a local log you can query yourself.

Free core · Apache-2.0

Your agent is about to make a payment. Give it rails first.

One command to install. No account required to try it.

$ brew install aor
# macOS, via Homebrew
 
$ curl -sf https://raw.githubusercontent.com/agentOnRails/agent-on-rails/main/scripts/install.sh | sh
# macOS or Linux, one-liner
 
$ aor init

Building from source works today too: go install github.com/agentOnRails/agent-on-rails/cmd/aor@latest