Open Agent Access
Consequential agents must carry papers. Open Agent Access (OAA) is the layer that decides whether an autonomous agent may read, may be refused, or must pay — on a published, machine-readable policy, with a signed record of every action.
Why a control layer?
robots.txt was built for crawlers that only read. Agents now act — they transact, book, and settle on a user’s behalf. That needs more than allow/disallow: it needs identity, a stated purpose, an enforceable decision, payment terms for actions that cost something, and an audit trail. OAA is that boundary — issued by the site, honoured by well-behaved agents.
It’s the governance counterpart to the rest of the toolkit: the scanners tell you whether agents can find and act on you; OAA is how you let them act on your terms.
Three decisions
- allowPublic reads — research, indexing, summarisation — with attribution by source URL.
- denyDisallowed purposes, e.g. AI-training use, are refused outright.
- chargeConsequential actions return an x402 payment-required envelope; settle to proceed.
How a request flows
- 1Present a passport. The agent sends its identity, purpose, and intended use (AA-Agent-ID, AA-Purpose, AA-Use).
- 2Policy decision. The policy engine evaluates the request against the published rules and returns allow, deny, or charge.
- 3Settle payment (if charged). For paid actions, the agent receives an x402 envelope and settles a micropayment on Algorand.
- 4Receive a receipt. The action returns a hash-chained receipt — an auditable record of identity, decision, and settlement.
Payments & receipts
Paid actions are priced in the policy (the Agent City demo charges $0.01 / request) and settled with x402 — USDC on Algorand (TestNet in the demo), with sub-3-second finality. Every governed action returns a hash-chained receipt: who acted, under what purpose, the decision, and the settlement — auditable after the fact.
OAA is also exposed as a live MCP server, so agent frameworks can discover and call governed tools directly rather than scraping the page.
See it work
Agent City is the live demonstration — autonomous agents present passports, receive allow / deny / charge decisions, settle real x402 micropayments on Algorand TestNet, and collect receipts, step by step. It’s where the whole toolkit comes together as agentic commerce.
Read the policy: /.well-known/agent-access.json.