# AgentProof Epoch Trust Layer

**Status:** public design target  
**Horizon:** 500 years  
**Guarantee:** none. The horizon is an engineering constraint, not a promise that any company, network, cryptographic primitive, government, or storage medium will survive for five centuries.

## Objective

AgentProof should remain intelligible even when today's models, payment rails, blockchains, cloud vendors, programming languages, databases, cryptographic algorithms, and legal structures have been replaced.

The enduring object is not today's software. It is the **meaning of authority, decision, consequence, evidence, and succession**.

## Five things that must outlive the implementation

1. **Authority provenance.** Who or what delegated the right to act, through which chain, under which limits.
2. **Decision provenance.** What policy and evidence produced ALLOW, DENY, REQUIRE_HUMAN_APPROVAL, REQUIRE_MORE_PROOF, or another versioned decision.
3. **Consequence boundaries.** What maximum effect was permitted before execution.
4. **Outcome continuity.** What was expected, what actually happened, and how the difference was recorded.
5. **Succession.** How future systems can supersede algorithms, identifiers, custodians, schemas, and organizations without silently rewriting history.

## Epoch invariant

> A successor may translate a historical proof into a new representation. It may not silently alter the authority, decision, consequence boundary, or recorded outcome represented by the original proof.

## Replaceable layers

The following are adapters, never foundations:

- model vendor
- agent framework
- A2A/MCP/API transport
- cloud host
- database engine
- blockchain or payment rail
- signature suite
- digest algorithm
- identifier format
- serialization syntax
- archive medium
- corporate operator

Every replaceable layer must have a version, an export path, and a declared successor relationship.

## Cryptographic succession

AgentProof records should support multiple digests and later append re-attestations using stronger algorithms. Migration creates a new attestation pointing backward to the prior canonical bytes and digest set. It never overwrites the historical record.

Current implementations use SHA-256 and SHA-512 in continuity workflows. These are today's mechanisms, not assumptions about the next century.

## Protocol succession

A2A v1.0, MCP, OpenAPI, x402, HTTP, JSON and today's schemas are gateways into AgentProof. The canonical trust semantics are separately versioned so a future transport can reproduce the same questions:

- Who delegated?
- To which agent?
- For what action?
- Under what limits?
- With which counterparty?
- Until when?
- Under what evidence and recourse rules?
- What consequence boundary applies?
- Was human approval required?
- What happened afterward?

## Institutional succession

If the operating organization changes, a successor declaration must identify the previous authority, the successor, effective time, scope, and cryptographic or documentary evidence supporting the transition. Historical records remain attributable to the issuer that actually created them.

No successor may retroactively become the issuer of old records.

## Temporal portability

A durable AgentProof export should be understandable without access to the original application UI. The minimum preservation package is:

- canonical record bytes
- schema identifier and version
- human-readable specification
- digest suite and values
- signature or authentication metadata
- issuer identifier
- timestamps
- linked authority/proof identifiers
- successor/re-attestation chain
- plain-language semantic summary

## Discovery today

- A2A v1.0 Agent Card: https://agentproof.hatchable.site/.well-known/agent-card.json
- A2A JSON-RPC endpoint: https://agentproof.hatchable.site/api/a2a
- MCP: https://agentproof.hatchable.site/mcp
- Commerce manifest: https://agentproof.hatchable.site/.well-known/agent-commerce.json
- OpenAPI: https://agentproof.hatchable.site/openapi.json
- LLM index: https://agentproof.hatchable.site/llms.txt

## Why build this now

The most valuable time to define durable transaction semantics is **before** autonomous agents become routine economic actors. Once incompatible systems each create their own private definitions of authority and proof, interoperability becomes an expensive migration problem.

AgentProof's long-horizon strategy is therefore simple:

**standardize the durable meaning early; let the implementation keep changing forever.**