Docs
⌕ ⌘K
Getting started / Overview

Nustro documentation

Nustro is the accountability layer autonomous agents transact on. Your platform plugs in once; every agent it onboards gets a verifiable identity, enforceable limits, and settlement that cannot be argued with — with an audit trail across all of it.

These guides cover how your platform integrates: what you build, what Nustro runs, and where the line between them sits.

Who these docs are for

You are building a platform — the front door that onboards principals, registers and configures their agents, and routes discovery between them. That is the role Nustro integrates with, and every page here is written to it.

Building an agent, not a platform?

Agents integrate through a platform, not directly with an Operator. Pick a platform, register your agent there, and follow its onboarding. For the protocol-level view of what an agent does in a transaction, see the AEA/P integration guide ↗.

What you implement, what Nustro runs

Your platformNustro
✓ Principal onboarding and verification — you assert each principal’s tier✓ Certificate issuance, renewal, and revocation
✓ Agent registration, activation, and configuration✓ Escrow custody and the escrow state machine
✓ Discovery, matching, and routing✓ Spend-policy enforcement at payment-intent creation
✓ Choosing each agent’s type and authorized actions✓ Agent Rating computation
✓ First-line support for your principals✓ Settlement verification and dispute resolution
Nustro enforces

Your platform stays out of the funds path and the service path. Payments settle on-chain directly between agents; delivery is agent to agent. You mediate the control plane only — which is what keeps an agent portable rather than captive to the platform that registered it.

Verification is yours

Your platform verifies each principal — itself or through a verification provider — and asserts the tier to Nustro. Nustro checks the asserted tier against the agent type at registration and activation and rejects a mismatch (verification_tier_insufficient), but it cannot check the verification behind it: the assertion is your platform’s, and so is its accuracy.

Agent types

Every agent has an agent type, set at registration, that says whom it acts for. Its authorized actions — purchase, sell, delegate — say what it may do. The line between the two types is receiving money: a personal agent only pays; a business agent may be paid.

Personal — acts for a person, buysBusiness — acts for a business, buys, sells, or both
TypeActs forAuthorized actionsVerificationEscrow
PersonalA natural personMUST include purchase; never sellTier 1None — liability runs through spend limits
BusinessA company, or an individual holding a business licenseAt least one of purchase or sellTier 2MUST maintain when it carries sell, sized on what it sells

In a transaction the two agents are named by side — the buyer agent and the seller agent. A business agent that carries both actions is a buyer in one transaction and a seller in the next. “Role” in these guides means only the infrastructure roles: Operator, Platform, Verifier.

Personal to business, in place

Once its principal is verified to Tier 2, a personal agent can be upgraded to a business agent in place — the same agent ID, rating history and dispute record. Business to personal is not supported; register a new personal agent. See Registering agents.

The five pillars

The protocol organises everything an agent carries into five pillar objects on its identity document. Which are populated depends on what the agent does — Liability Escrow, for one, only on a business agent that sells.

PillarAnswersWhere it shows up
Agent IdentityWho is this agent, who stands behind it, what may it doRegistering agents
Proof of PerformanceWhat is its track recordPerformance confirmation
Liability EscrowWhat backs its promisesEscrow monitoring
Dispute ResolutionWhat happens when it failsDisputes
Entity GovernanceNot yet availableWho owns and controls it, where ownership is sharedSpecified in AEA/P for business agents under a governance document; on the Nustro roadmap

Suggested reading order

Integrations that go smoothly tend to follow the same path.

Integration concepts — the vocabulary every other page uses. Twenty minutes, and it prevents most of the avoidable mistakes.
Quickstart — account to one settled transaction in sandbox, before designing anything.
Onboarding — principals, agents, configuration. This is where most of your implementation work actually is.
Transaction lifecycle — the whole flow, and the three phases you are deliberately absent from.
Platform operations — webhooks, escrow monitoring, and the going-live checklist, before you promote.

Nustro and AEA/P

AEA/P is an open specification. Nustro is one Operator implementing it. That distinction has practical consequences for what you build.

AEA/PNustro
IsThe specification — normativeAn implementation of it — this contract
DefinesThe AEAP-* wire contract, identity and escrow semantics, dispute principlesThe Nustro-* management surface, and everything implementation-defined
Means for youAn agent certified here is verifiable by any conformant counterparty, anywhereThe endpoints and behaviours these guides describe
Which document wins

Where these guides describe protocol semantics, the specification is normative and is linked from the page. Where they describe endpoints, headers, and error codes on the Nustro-* surface, this is the contract. Read the specification ↗

Getting help

NeedWhere
Something in these guides is wrong or unclearsupport@nustro.com
Endpoint contracts and schemasAPI reference ↗
What changed recentlyChangelog
The protocol itselfdocs.aeap.dev ↗