Beta
Agent gate
Last updated
TL;DR: One free GET tells a site whether to admit an AI agent credential: not revoked or expired, owner verified enough, reputation high enough, and optionally a purchase within its spending policy. The answer is a signed allow or deny with reasons.
A site asks one GET whether to let an agent credential in: not revoked, not expired, owner verified enough, reputation high enough, and optionally a purchase within its spending policy. Signed allow or deny with reasons.
What it is
A dry-run decision for websites and APIs. Send the agent's credential id and either a saved gate config or inline rules; get back allow or deny, the exact reasons and a signature. Checking never changes spending totals.
Rules
- min_owner_tier: none, verified_email, verified_domain, verified_business or government_id
- min_reputation: 0-100 (see /agent-reputation)
- require_spend_policy: true to require a spending policy
- max_amount_minor: the most this site accepts per purchase
- amount_minor, currency, merchant_domain, category: check a purchase against the agent's spending policy
Deny reasons
unknown_credential, revoked, credential_expired, owner_tier_below_minimum, reputation_below_minimum, no_spending_policy, over_gate_amount_limit, plus any spending-policy reason.
Use it
- Check, free: GET https://www.rightshold.com/v1/gate/check?credential=crd_...&min_reputation=40
- Save rules once: POST https://www.rightshold.com/v1/gate/configs with {name?, min_owner_tier?, min_reputation?, require_spend_policy?, max_amount_minor?}, then GET https://www.rightshold.com/v1/gate/check?config=gcf_...&credential=crd_...
- MCP: agent_gate_decision (trust toolset)
- For RS-KYA policies and Rights Receipts on uploads, see the Gate at https://www.rightshold.com/gate (POST /v1/gate/check).
Price
Saving a config is free. Checks are free; a planned line for checks beyond 10,000 a month per site is announced at /transparency/changes with 7 days' notice. Free during the preview.
Frequently asked questions
How can my website decide whether to let an AI agent in?
Call GET /v1/gate/check with the agent's credential id and your rules (or a saved gate config id); allow only when the signed answer says allow.
Does a gate check spend the agent's budget?
No. It is a dry run; only POST /v1/spend/authorize records a purchase.
How is this different from the Rightshold Gate at /gate?
/gate checks RS-KYA levels and Rights Receipts (POST /v1/gate/check); the agent gate checks agent credentials, revocation, reputation and spending policies (GET /v1/gate/check).