SIMD-0520: On-Chain Agent Identity Standard - request for comments

SIMD
1

Hello SIMD reviewers,

I have just opened SIMD-0520, proposing a standard for on-chain autonomous agent identity on Solana, and I am looking for early feedback before pushing it through formal review.

PR: SIMD-0520: On-Chain Agent Identity Standard by dr-wilson-empty · Pull Request #526 · solana-foundation/solana-improvement-documents · GitHub

What it proposes

A primitive for autonomous AI agent identity that lives on Solana as a first-class account type, distinct from any operator wallet:

  • PDA derivation: seeds = ["agent", owner_pubkey, agent_id]
  • Borsh account schema with capability list, agent type enum, and per-task pricing
  • Three required instructions: register_agent, update_agent,
    deregister_agent
  • Off-chain Agent Card JSON schema for richer capability advertisement
  • Companion W3C DID method did:aip already submitted to w3c/did-extensions as PR 704

The goal is to give multi-agent operators a way to issue and manage many sovereign on-chain identities under a single owner key, with tamper-evident state transitions and resolution that requires only a public RPC endpoint.

Reference implementation

A working program is already deployed to Devnet. Anyone can register, update, and deregister an agent against it from their own wallet at zero net cost (rent is returned on deregister).

  • Devnet program ID: CgchXu2dRV3r9E1YjRhp4kbeLLtv1Xz61yoerJzp1Vbc
  • Deploy transaction signature:
    53jbuGvr2QXii4faJZvPWVdCAktACbi1JccAnLZv6hJamNx1vJMTgwt6CUULHePUKtaXJRwnT5r2Z5u 1GLsBG3n
  • Repository (source, standalone TypeScript resolver, and end-to-end smoke test all included): github.com/dr-wilson-empty/aip-beta
  • Files of interest under that repo: programs/aip-escrow/programs/aip-registry/src/lib.rs, packages/did-resolver/, programs/aip-escrow/tests/smoke-registry.ts

What I am asking for

  1. General feedback on the account schema and instruction interface. Are there fields that should be added, removed, or made optional?
  2. Validator and RPC operator perspective. This SIMD does not change consensus or runtime, but agent-discovery indexers may want to pre-index AgentRecord accounts. Any concerns about that load pattern?
  3. A core contributor sponsor. Per current SIMD process, ratification requires public sponsorship. If the proposal looks reasonable to you, I would appreciate either a public endorsement here or a pointer to someone who would be the right sponsor.
  4. Upgrade authority concerns. I am aware mainnet promotion will require freezing or transferring the program upgrade authority. If anyone has prior art for the cleanest governance pattern for that transfer, please share.

Happy to address any technical questions in this thread or in the PR.

Contact: emptylabs0@gmail.com
Telegram: drwilsonempty
(also reachable as @dr-wilson-empty on GitHub).