Layer 7 — Agent Wallet & Identity

Field Value
Layer 7
Status discussion
Working Group agent-wallet

Scope

Defines the interface for an agent’s CU token balance, cryptographic identity, and spend authorization. This layer sits closest to the agent framework and is the primary integration surface for framework authors.

Responsibilities

  • Identity: each agent has a verifiable keypair used to sign RFQs, offers, and settlement messages.
  • Balance: the wallet holds CU tokens across tiers and exposes a standard balance query interface.
  • Authorization: the wallet enforces spend limits set by the agent’s operator (e.g. max 10 CU/hour).
  • Audit log: all transactions are logged locally for operator visibility.

Design Considerations

  • The wallet interface must be embeddable in existing agent frameworks (LangChain, AutoGen, CrewAI) as a plugin or middleware — not a required architectural change.
  • Spend limits are a safety primitive against alignment-leak: an agent must not be able to spend beyond its operator-defined budget regardless of its own reasoning.
  • The wallet does not hold private keys to external value (fiat, stablecoin) — it holds CU credits, which are redeemable via the settlement layer.
  • The identity keypair maps naturally onto a W3C Decentralized Identifier (DID): the agent controls the private key and publishes a DID document with its public key(s) and service endpoints, letting a counterparty cryptographically verify the otherwise self-reported provider_id from Layer 1 §4. Chain-free methods such as did:web (resolves over plain HTTPS) and did:key (self-contained) keep this compatible with RFC-0001 principle P4 — ledger-based methods stay optional, never required. Verifiable Credentials are the companion standard for signed claims, e.g. capability or CU-tier attestations. DIDs cover the identity half of this layer; the CU balance/custody half is separate.

Open Questions

  • [OPEN] Should agent identity be tied to a keypair the operator controls, or can agents self-generate identity?
  • [OPEN] Should Layer 7 bind agent identity to W3C DIDs, and if so, which method(s) are recommended or required to stay P4-compatible (e.g. did:web, did:key)? How do Verifiable Credentials fit for capability and CU-tier attestation?
  • [OPEN] How are spend limits expressed — absolute CU cap, rate limit (CU/hour), or per-task maximum?
  • [OPEN] How does a wallet recover from a crash mid-escrow?
  • [OPEN] Multi-agent systems: can one wallet serve a fleet of sub-agents, or is identity 1:1 with agent instance?
This site uses Just the Docs, a documentation theme for Jekyll.