How to Charge AI Agents for Access to Your Site
Published November 14, 2026 · Reviewed November 2026 · By the Ingenious AI team
The short answer
Charging AI agents, in AGENT-POLICY v0.1, means one concrete step: set commercial_terms to paid-tier on the rules where access has a price, and have a real paid tier behind the signal — details, prices, and onboarding reachable through your owner.contact. The field carries no amounts and runs no payments; v0.1 defines no negotiation or payment protocol. What it does is put a price tag's existence in the standard place, so agents and operators can find it without asking.
What does paid-tier actually declare?
It declares that this rule's access runs through a paid offering the owner operates. Read by an agent, it converts "may I work here?" into "my operator has a commercial step to complete". Read by an operator, it is a signpost: a tier exists, its details live with the owner, and the contact in the file's header is the door. The declaration's honesty requirements are symmetrical with the rest of the format — as the commercial terms guide puts it, a paid-tier signpost pointing at no tier, no prices, and no respondent is the one genuinely bad use of the field.
Decide what the meter runs on
Real access tiers price one of three things. The data: catalogue, corpus, or archive access, priced like the licensing it resembles — the model for publishers and catalogue owners. The pace and volume: a free tier at reading speed, a paid tier at working speed — the store sample's shape, where the paid rule is also the 60-requests-per-minute rule. The job: transactional access — the right to book, reserve, or transact under agreed terms — priced per relationship rather than per request. Most owners blend the first two. What matters is that the tier behind the signpost has a shape an operator can evaluate in one email exchange: what is included, at what pace, for how much, starting how. Everything beyond that first exchange is ordinary B2B contracting, which the format wisely leaves alone.
Build the smallest real tier
A tier does not need a billing platform to be real. It needs: a name, a scope (which agents, which actions, what pace), a price or a pricing conversation with a floor you know, and a human who answers the contact address within a businesslike interval. Plenty of working tiers are a one-page sheet sent in reply to the first enquiry — the policy's job is only to start that enquiry from the right agents, pre-qualified by the rule they read. Owners who already run API pricing or data licensing have the sheet written; the policy work is pointing the right rules at it. Owners starting from zero should begin with contact terms — negotiate first — and graduate rules to paid-tier when the sheet exists, rather than declaring a tier that is still an intention.
Keep the free parts free on purpose
Charging works best as a boundary, not a blanket. Reading and quoting are most sites' shop window; pricing them usually costs more demand than it earns. The durable pattern across real policies: free for the verbs that market you, paid-tier for the verbs that cost you or arm competitors — bulk access, working-speed catalogue use, premium data — and contact for strangers. The fields cooperate: a rule can permit read while its sibling disallows crawl-full-site except on the paid tier's separate arrangement, with data-use terms spelling the after-life for each. Price the value, publish the boundary, answer the door.
Worked example: a trade directory builds a two-step ladder
Fictional example. BuildRegistry, a fictional trade directory listing 48,000 contractors, runs 3 term levels. Every named agent reads listings free at 20 requests per minute — the directory's public face. Verified data — licence status, insurance dates, review histories across 12 fields per listing — sits behind paid-tier: 2 agent operators signed in the first quarter at a flat monthly tier the owner quotes by email, receiving a raised 120-per-minute cap in a revised rule as the visible token of the deal. Bulk export remains contact-only, negotiated per case, and was declined twice. The owner's accounting is simple: the free tier markets the directory, the paid tier monetises the verification work the directory actually spends money on, and the policy file carries both price signals in 3 lines.
The charging checklist
- Find the value first. Name what agent access costs you or earns competitors — that is what the tier prices.
- Write the tier sheet. Scope, pace, price basis, and how to start — one page an operator can judge.
- Set paid-tier only where the tier is real. A signpost to nothing is worse than
contact. - Keep the window free. Leave reading and quoting on
freeterms unless your whole product is the content. - Staff the door. Every paid-tier rule routes to
owner.contact; the tier sells at the speed that inbox answers.
Keep reading
Conditional Access: Making Agents Contact You First · How to Block All AI Agents (and Say So Clearly) · Agent Policy vs robots.txt, Compared · Spec samples · Conformance · Pricing · All guides
Frequently asked questions
Does AgentGate process payments for my tier?
No. AgentGate's own pricing is for its generator, validator, hosting, and batch tooling. Your access tier is your commercial arrangement with agent operators, declared through your policy's commercial_terms and conducted through your owner contact. AGENT-POLICY v0.1 defines no payment protocol and moves no money.
What should a paid data tier cost?
The format takes no position, and neither does this guide — pricing is a property of your data's value and your market. The workable pattern is to price what the tier replaces: a licensing fee for data, a subscription for pace and volume, a per-relationship fee for transactional rights. Start where one email exchange can close the question.
Can I charge one agent and not another?
Yes. Terms are per rule: a favoured partner can hold free terms at high pace while the catch-all faces paid-tier or contact. Per-agent pricing is the normal shape of B2B access, and the format's rule structure exists to express exactly these differences.
Should I switch a free rule to paid-tier later?
You may — the file is yours to revise, with last_updated moved and the change validated. Treat it as a commercial change, not an edit: notify operators who built on the free terms, because their agents cached a file that said free. Policy changes with paying counterparties deserve the change-history habits in the versioning guide.
What if agents will not pay?
Then the market has told you something about the tier, not about the format. The policy still does its other work — declaring actions, fences, and data-use terms — and the contact value keeps a negotiation channel open. Many owners conclude free-with-fences is their best commercial answer; the field's job is to let them say whichever answer is true.
Generate a policy — free · Validate a policy · What an agent access policy is