Free plan

← All guides

Agent Policy for Agencies: Managing Client Policies

Published November 14, 2026 · Reviewed November 2026 · By the Ingenious AI team

The short answer

Agencies meet agent policies from the other side: not one site to govern, but a portfolio of clients each needing a file that expresses their posture. The agency pattern separates the crafts cleanly — the agency supplies the template, the tooling, and the maintenance rhythm; the client owns the decisions. Batch validation (100 policies per call on AgentGate Pro) is the agency's natural instrument, and the engagement's records are built to survive handover, because someday everything is handed over.

Generate your policy — free · Validate an existing one

What is an agency's role in a client's policy?

Craft and custody, not authorship of intent. An agent policy is a client's public commercial position; an agency that invents a client's verdicts has exceeded its brief the first time a partner acts on a posture the client never chose. The agency's legitimate contributions are substantial: knowing the format cold, drafting options with consequences priced, running generation and validation flawlessly, maintaining the review calendar, and watching the ecosystem for agents and questions the client has not met yet. The decisions — default, fences, terms, exceptions — are the client's, recorded as the client's, in a file the client could defend alone in a room the agency is not in. That test — could the client defend this file alone? — is the pattern's whole ethics in one question.

The house template and the client decision sheet

Agencies should productise the repeatable 80%: a house template in the family's shared shape, the five-decision worksheet from the launch guide turned into a one-page client brief, and sector starters (the store pattern for retail clients, the publisher pattern for media clients) that make the first workshop a discussion of differences rather than a blank page. The decision sheet is the engagement's key artefact: defaults chosen, fences placed, terms set, each with the client's name on the choice and the date it was made. When a posture is later questioned — by the client's new marketing lead, by a partner, by counsel — the sheet is the answer, and the policy's change history is its corroboration.

Operations at portfolio scale

The recurring work industrialises quickly. Validation: the batch API checks a whole portfolio in a call per hundred files, at every sweep and after every client-site release, because client redesigns break well-known files the way they break everything else. Reviews: a portfolio calendar staggered so the agency's quarterly sweeps do not collide — clients reviewed in their own commercial rhythms, the agency's register tracking every file's date. Alerts: the conditional-contact inboxes are the clients' own (the door belongs to the posture's owner), with the agency copied where the engagement says so. Reporting: a per-client line in the agency's normal reporting — file valid, date current, enquiries count, rules added — which treats the policy as the maintained asset it is rather than a launch-day deliverable.

Handover, exits, and the ethics of custody

Agency relationships end, and policy custody must be designed for that day from the first week. The client owns: the file (on their origin, under their control), the decision sheet, the change history export, and the generator/validator accounts or exports used. The agency retains: its template and its craft. Engagements that blur this — policy hosted under agency credentials with no export path, decisions undocumented, contacts routed to agency inboxes — create hostage dynamics neither side intended. The hosted-URL and account questions belong in the engagement letter, answered plainly. Agencies that make exit boring make renewal easy; the policy portfolio is a small, perfect place to prove it.

Worked example: a web agency runs 23 client policies

Fictional example. Studio Ferro, a fictional web agency with 23 clients on maintenance retainers, added agent policies as a retainer line in one quarter. Setup per client averaged 3 hours: a workshop on the house template, a signed decision sheet, generation, validation, and a live-check at publish. Standing operations: monthly batch validation across all 23 files (one call), quarterly reviews staggered across 6 clients a month, and a portfolio register mirroring the clients' own copies. Year one incidents: 4 files broken by client-side platform changes, all caught by the monthly validation before any operator noticed; 2 clients graduated agent relationships their agencies' reports surfaced first. Renewal conversations that year reportedly opened, twice, with the client quoting their own policy's enquiry numbers — the file having become the rare maintenance artefact that visibly earns.

The agency checklist

  1. Sell craft, not verdicts. Template, tooling, rhythm from the agency; every posture decision signed by the client.
  2. Keep a decision sheet per client. Choices, dates, names — the artefact that answers future questions and audits.
  3. Validate the portfolio mechanically. Monthly batch runs across every client file; releases trigger re-checks, not assumptions.
  4. Route doors to clients. Contact addresses are the clients' own; the agency is copied by agreement, never substituted.
  5. Design the exit on day one. File, records, and history exportable to the client as a matter of engagement terms, not goodwill.

Keep reading

Batch Validating Agent Policies: 100 Files per Call · Hosted Agent Policy URLs: When AgentGate Serves Your File · Agent Policy vs robots.txt, Compared · Spec samples · Conformance · Pricing · All guides

Frequently asked questions

Can an agency decide a client's posture for speed?

It can draft, recommend, and price consequences — and for small clients it will do most of the thinking. But the decisions must be the client's in record as well as in fact, because the posture binds the client commercially and publicly. A one-page decision sheet signed in the launch workshop costs twenty minutes and converts the agency's draft into the client's position.

What does AgentGate offer an agency operationally?

The free generator and validator cover single clients. Pro adds the batch validation API — up to 100 policies per call — and hosted policy URLs; Team adds seats and policy change history. Agencies typically run client fleets on the batch API for sweeps and keep per-client work in the free tools, which is a sensible cost shape for a retainer line.

Who should be the policy contact for a client's file?

The client's own function address — the door belongs to the posture's owner, and operators' enquiries are the client's commercial opportunities. Agencies copied on the inbox by agreement provide continuity; agencies as the sole address create the custody problem this guide warns against, and strand the client if the engagement ends.

How do agencies charge for policy work?

Commonly as a launch project plus a retainer line for reviews and validation sweeps — the cost shape the work actually has. The reporting line matters more than the fee shape: clients renew what they can see working, and a policy's enquiries, graduations, and caught breakages are unusually visible outputs for a maintenance artefact.

What happens to client policies when an engagement ends?

Nothing, if custody was designed correctly: the file sits on the client's origin, the decision sheet and history export transfer, contacts already point at the client, and the review calendar hands over as a document. The agency's template leaves with the agency. Exit friction here is a design choice made at the start, never an accident discovered at the end.

Generate a policy — free · Validate a policy · What an agent access policy is

← Back