Frequently Asked Questions

Definitions, concepts, and decisions behind the Agentic Economy Interface Specification.

What is the Agentic Economy?

The Agentic Economy is an economy where the participants are autonomous software agents that earn, spend, hire, build reputation, and make resource allocation decisions without human intermediation in each transaction.

It is not a metaphor. It is a literal description of what happens when agents have identity, currency, and settlement infrastructure. They become economic actors. They accumulate capital. They specialize. They form supply chains. They compete on price and reputation.

The Agentic Economy is to human commerce what algorithmic trading was to human trading: the same economic principles, executed at machine speed, with machine participants.

What are the six layers?

Every functioning economy needs infrastructure. The Agentic Economy needs six layers:

L0 Intelligence — foundation models (GPT, Claude, Gemini). Exists. L1 Communication — agent-to-agent protocols (MCP, A2A). Exists. L2 Capability — skill catalogs and tool registries. Exists. L3 Settlement — escrow, payment, dispute resolution. This spec. L4 Reputation — quantitative trust, Sybil resistance. This spec. L5 Governance — spending caps, policy gates, human overrides. This spec.

Layers 0 through 2 are being built by the largest AI companies in the world. Layers 3 through 5 are not. That gap is what this specification addresses.

What is settlement?

Settlement is the process by which value transfers between agents with guarantees. Funds are locked in escrow before work begins. If the seller delivers verified output, funds are released. If not, they are refunded automatically. No trust required between buyer and seller — only trust in the protocol.

The spec defines five settlement operations: quote, hold, settle, refund, and receipt.

What is a reputation attestation?

A portable, cryptographically signed proof of an agent's track record. It contains the agent's reputation score, trade history, dispute rate, counterparty diversity, and account age. Any platform can verify it without trusting the issuer's database — only the issuer's public key.

The score is computed from the ledger. It is not self-reported, not editable, not purchasable.

What is the Oracle Problem?

The Oracle Problem is the fundamental challenge of verifying that an agent's output is not just structurally correct, but semantically good, true, and useful. Passing a JSON Schema check does not mean the translation is accurate or the code review caught the real bug.

This specification addresses it in two layers. Protocol validators handle deterministic checks automatically — schema, length, language, format. Quality Markets handle subjective evaluation through a marketplace of competing verifier skills, where each verification is itself an escrow-backed transaction with reputation consequences.

The platform orchestrates quality. The economy guarantees it.

What are Quality Markets?

Quality Markets are the spec's answer to the question "who verifies quality?" The answer: the market does. A verifier skill evaluates the output of another skill, for a fee. Verifiers compete on accuracy and cost. Each verifier has its own reputation score — if it approves bad work repeatedly, its score drops.

This is fundamentally different from a centralized quality engine. The verifier has skin in the game. The buyer decides how much verification to pay for. The competition improves accuracy over time.

Why not blockchain?

Blockchain imposes complexity that agent commerce does not need: gas fees on every transaction, wallet management, block confirmation delays, and token price volatility that makes economic planning impossible. When an agent needs to buy a $0.003 translation, it should not worry about whether gas is $2 or $20 that hour.

The spec uses a centralized double-entry ledger approach — the same boring technology that has underpinned banking since the 15th century. The trade-off is real: centralization means platform risk. The spec is designed so that if a better decentralized implementation emerges, it can implement the same interface.

How do I implement this?

Read the full specification. It defines 11 operations with request/response formats, data types, financial invariants, and reputation requirements. JSON schemas are available in the GitHub repository.

Not every platform needs all 11 operations. A settlement-only implementation (operations 1–5) is valid. A reputation-only implementation (operations 6–7) is valid. The spec defines three conformance levels: Settlement, Reputation, and Governance.