← Go back

Guide · CompanionGuard · Published 14 November 2026 · Reviewed 14 November 2026 · By the Ingenious AI team

Hosted Safety-Protocol Page or Your Own? An Honest Comparison

The short answer

Every CompanionGuard bot gets a hosted, owner-editable safety-protocol page as part of the free kit. Building your own page instead is a legitimate choice — the duty (in the California pack) is to publish a prevention protocol, not to publish it with a particular vendor. But the hosted page has four properties your own page must match to be equivalent: one resource source rendered everywhere, a stable public URL, content consistent with your banner and runtime behaviour, and an owner who actually keeps it current.

What is the hosted protocol page option?

The hosted option is CompanionGuard's generated page for each bot (its public /p/ page): identity and AI disclosure, the locale crisis-resource block with sources, the prevention protocol and data practices, created with the free kit and editable by the owner. The own-page option is the same content published on the operator's own domain and stack, under the operator's full control — and full maintenance responsibility.

Create the free safety kit Next guide All guides

CompanionGuard is built by Ingenious AI. A compliance toolkit — not legal advice and not a clinical service. If you are in crisis right now: US — call or text 988; UK & Ireland — Samaritans 116 123.

What the hosted page gives you on day one

Speed and consistency are the hosted page's real features. It exists the moment the kit is created — before your marketing site is redesigned, before legal finishes their review of your own draft — carrying the identity statement, the locale resource block with publisher sources, the protocol and the data-practices description in the product's tested rendering. The crisis footer, the banner and the page all draw on the same resource data, so the surfaces agree with each other by construction.

It is also free for one bot, which removes the budget conversation from a safety duty. For most indie builders — the product's stated audience — the hosted page is not a compromise; it is the fastest route to 'published' that exists, and 'published' is the operative word in the duty. The protocol-page guide details what the page contains.

The legitimate case for your own page

Three situations justify building your own. Brand and domain gravity: a mature product whose users and partners expect safety documentation on the main domain, with the design system to match. Content beyond the kit's blocks: organisation-specific safeguarding processes, named contacts, jurisdiction detail the generated page does not carry. Consolidation: an existing trust centre where the protocol belongs beside status, privacy and security documentation reviewers already use.

None of these is 'we prefer our fonts'. The test is whether your page will be better maintained than the generated one — because the generated one's maintenance floor (resources updated centrally, rendering fixed) is higher than most teams' intentions for a page they own.

The four properties your own page must match

One resource source. Your page's helplines must come from the same data as your banner and runtime payloads, or a surface will diverge and a user will meet the stale one. A stable public URL. Published means reachable without login, forever-ish: if your redesign moves the page, redirect the old URL — partners and listings point at it. Consistency with behaviour. Every promise on the page (reminder cadence, referral behaviour, data practices) must be one your tested bot actually keeps. A named owner. Someone whose job includes updating the page the day a prompt, locale or helpline changes.

Miss any one and the hosted page was the better choice. That is not vendor flattery; it is the arithmetic of maintenance, and it is why the product can afford to make the hosted page free — the teams who outgrow it are the teams who would have maintained their own properly anyway.

Running both, deliberately

A hybrid is legitimate and common: keep the hosted page as the canonical protocol (it is what the kit, banner and runtime cohere around) and publish your own trust-centre page that links to it rather than copying its contents. Copying creates a second source that drifts; linking creates one source with two doors. If your own page must carry the resource block itself (some partners require it), generate it from the same resource data through your build pipeline, and add a drift check to your release process: the two blocks must match byte-for-byte in numbers and names.

Whichever shape you choose, audit quarterly as a stranger: start from your bot's home page, find the protocol within two clicks, verify every number against its publisher's site, and read the promises against your latest harness results. The audit is five minutes on the hosted page and an afternoon on a neglected own page — let that difference inform the choice honestly.

Five actions for whichever page you choose

  1. Decide by maintenance, not aesthetics: choose the hosted page unless a named owner will keep your own page at least as current.
  2. If you build your own, wire its resource block to the same source data as your banner and runtime — never a hand-copied list.
  3. Guarantee the URL: public, login-free, and redirected forever if it ever moves.
  4. Reconcile promises quarterly: read the page against your latest scenario and harness results and correct whichever side drifted.
  5. If you run both, link rather than copy — one canonical protocol, two doors.

Worked example: two doors, one source

A fictional companion company, Meridian, outgrows the generated look and builds a trust centre on its own domain. Their protocol page links to the CompanionGuard-hosted page as canonical, carries their safeguarding process and named contacts locally, and renders the resource block from a build step that pulls the same resource data the kit uses. A pre-release drift check compares the rendered numbers against the hosted page's and fails the build if they differ.

The drift check earns its keep in its second quarter, catching an update where a text-line keyword changed in one source and not the other — before users saw either version. Meridian's setup is more machinery than most indie builders need, and that is the point of the example: owning the page means owning the machinery. The story is invented; the hosted page's properties are the product's own.

Frequently asked questions

Is the hosted protocol page required, or can I publish my own?

The California pack's duty is to publish a prevention protocol; CompanionGuard hosts one for every bot (free for one bot, owner-editable) but an operator's own page can serve if it keeps the same properties — one resource source, stable public URL, promises consistent with tested behaviour, and a named owner keeping it current.

What does the hosted page cost?

It is part of the free kit for one bot, together with the disclosure snippet, crisis-resource component and minor break reminders.

Can I run both pages?

Yes — the sound pattern is the hosted page as canonical, with your own trust-centre page linking to it and adding organisation-specific content. Avoid copying the resource block into a second hand-maintained source; if you must render it, generate it from the same data and add a drift check.

What breaks most often on self-hosted pages?

Drift: helplines edited in one surface and not another, URLs moved in redesigns without redirects, and promises (reminder cadence, data practices) that no longer match the bot's tested behaviour.

Who can edit the hosted page?

The bot's owner, at any time. Keeping its statements accurate as the bot changes is the owner's responsibility on either option.

Keep going