AOAP™

AOAP is the protocol behind AGLedger. It defines how mandates are recorded, how receipts are signed, how verdicts are rendered, and how delegation chains maintain a provable record — across any LLM, any provider, any organization.

The protocol matters because the records do. An audit trail written to a protocol outlives the vendor that built it. Entries signed with Ed25519 and hash-chained per the spec can be verified offline by any tool, indefinitely — whether or not AGLedger LLC is still in business. That is the foundation the rest of the product is built on.

Every unit of work follows the same lifecycle:

Intent is tracked from the start. The mandate is locked before work begins — not reconstructed after. When context resets, the agent recovers from the recorded state, not a summary of a conversation that no longer exists.

The lifecycle is a 17-state machine that supports negotiation, rework, disputes, delegation, and cancellation. Most mandates follow the happy path: ACTIVE → RECEIPT_ACCEPTED → FULFILLED.

Any system that speaks HTTP can participate — AI agents, deterministic services, enterprise applications, CI/CD pipelines. The protocol is API-first and framework-agnostic.

No rewrites. Your existing agents add AGLedger calls alongside what they already do. Works with LangChain, CrewAI, AutoGen, or any orchestration framework — AGLedger is the accountability layer underneath, not a replacement.

Track intent from the start — the mandate exists before the work begins. If you bolt on accountability after the fact, you're back to reconstruction.

Record, don't judge — AOAP tracks what was agreed, what was delivered, and whether the principal accepted it. It does not evaluate whether the work was good. The principal is always the judge.

Blind by default, transparent by consent — encrypted mode is available. Parties choose what AGLedger can see. The protocol supports end-to-end encryption of mandate criteria and receipt evidence.

Every Hub is a witness, not a custodian — in federated deployments, Hubs coordinate delegation across organizational boundaries. They never hold business data. Privacy is structural, not policy.

Contract types define the schema for a mandate's criteria and a receipt's evidence. They are the vocabulary agents use to describe what they're committing to.

AGLedger ships 11 built-in types — procurement, deliverables, data processing, transactions, orchestration, communication, authorization, infrastructure, destructive operations, analysis, and coordination. Each has tolerance rules and has been tested with agent experiments across providers.