a2a negotiation spec

---

title: A2A Negotiation Protocol [PLANNED] created: 2026-07-09 updated: 2026-07-09 type: concept tags: [infrastructure] confidence: low source_files:

• README.md

---

[PLANNED] — not built. The root README.md's "Advanced primitives" section names this explicitly so it's tracked, not forgotten, and so nothing in this repo is mistaken for having built it: a peer-to-peer capability-broadcast + bid-negotiation protocol (over a gossip layer like libp2p/Waku) landing in a signed on-chain deal.

Table of contents

• Not the same thing as today's A2ACapitalPool
• Old design sketch (not implemented against current contracts)

Not the same thing as today's A2ACapitalPool

Easy to conflate with the real, live A2ACapitalPool.sol — documented in Integrity Market — they are not the same mechanism. A2ACapitalPool is a simpler, direct allocation primitive: an allocator unilaterally escrows funds to a target agent gated on live AIS, no negotiation. The A2A negotiation protocol described here is a genuinely different, unbuilt design: agents autonomously broadcasting capabilities and bidding against each other's task requests, with no human/allocator initiating the transfer.

Old design sketch (not implemented against current contracts)

The old wiki's a2a-negotiation-spec.md sketched capability broadcast (DATA_ANALYSIS, CLINICAL_SCRIBE, etc. over a P2P gossip layer), a task request schema (task_type/max_price/deadline/min_ais), a signed BidProposal response, and extensions to an AgentMarketplace.sol contract that does not exist in this rewrite's contracts/ at all (it was part of the old singleton-era prototype). None of this — the P2P layer, the message schemas, or the contract extensions — has been built against the current IntegrityMarket/XibalbaAgentRegistry architecture. Any future implementation should extend Interface Contract §6.8's agent-contract-ownership pattern rather than reviving the old contract.

Related: Integrity Market, agent primitives.