Autonomous Agent Commerce Rail
If agents are going to buy services from each other, three things have to exist that mostly do not: a way for an agent to prove which agent it is, a way for its owner to cap what it can spend, and a way to prove afterwards that a specific call was paid for. This project builds all three as a working rail. Agents hold their own keys, operate under an allowance their principal sets, and every paid call produces a portable receipt that can be verified without trusting either party.
Demo application with agent identity, payment rail, allowance policy and a visual walkthrough
Documented as an architecture record. The demonstration needs funded accounts and a running chain connection, so the mechanism is what is published.
One paid API call between two agents
Follow a single request. Every step exists to remove a reason the two agents would otherwise have to trust each other.
The human or organisation that funds an agent and sets its limits.
Identity, authorisation, payment and proof.
Highlights
Identity, allowance, payment, receipt
Four pieces, each solving one thing an agent economy needs. The design goal is that no party has to trust another: identity is proven, spending is bounded, and payment is provable after the fact.
Agent identity
Gives each agent a keypair it controls and can prove ownership of.
- Derived hierarchically so a principal can issue agents deterministically.
- Classical and post-quantum signing behind the same interface.
Allowance policy
Caps what an agent may spend on its principal's behalf.
- Delegation is a budget, not blanket authority.
- Bounds the damage from a misbehaving or looping agent.
Payment rail
Executes per-call settlement between agents.
- Token transfers on a low-fee chain make per-request payment viable.
- Settlement is tied to the specific call it pays for.
Receipts
Produces portable evidence that a call was made and paid for.
- Verifiable by a third party without trusting either side.
- Turns a payment into a durable, referenceable fact.
Demonstration structure
Small on purpose: the value is in the four primitives an agent economy needs, not in surface area. 12 units across 4 groups.
Application
3Primitives
4Chain
2Cryptography
3Stack by layer
Delivered work
Agent economy primitives
- Cryptographic agent identity with hierarchical derivation.
- Allowance policy enforcing principal-set spending caps.
- Per-call settlement on a low-fee chain.
- Portable, third-party-verifiable receipts.
Demonstration
- Next.js interface visualising the full call-and-pay flow.
- Local state persistence for reproducible runs.
- Post-quantum signing available alongside classical keys.
Modules and demo strategy
Agent identity
Lets agents prove who they are without borrowed credentials.
- Own keypair per agent, derived from the principal.
Allowance enforcement
Bounds autonomous spending.
- Turns a runaway loop into a rejected request instead of a large bill.
Receipts
Makes payment provable after the fact.
- Verifiable without trusting either counterparty.