# Permission-as-a-Service — the Rad family category

> The platform thesis behind the Rad family: AI you can trust BECAUSE it
> refuses. Every product expresses one verb in spirit — request permission
> for an action, classify it by reversibility, and HOLD anything that can't
> be undone for a human. The agent acts; the human decides; always.
> "One valve, every surface."

## What this is
Permission-as-a-Service (PaaS, the pun is intentional) is the category the Rad
family defines. The cloud rents compute-as-a-service and storage-as-a-service;
the agent era needs one more layer — PERMISSION-as-a-service: the deterministic
mechanism that decides whether a consequential action gets to happen at all.

This page (radlife.ai/permission-as-a-service) is the developer- and investor-
facing positioning of that category. It is distinct from the RadLife umbrella
home page (radlife.ai), which tells the product story and roll-call. This page
is the platform thesis + the shared verb + the agent-discoverability wedge.

## The shared verb (ILLUSTRATIVE — not a shipped API)
Conceptually, every Rad product speaks one verb:

    requestPermission(action, input) -> { go | held | never }

- go    — reversible / read-only / benign. The agent does it autonomously.
- held  — irreversible or consequential (money, deletion, binding commitment,
          identity-of-record change, publish). Held for a human's one-tap yes.
- never — an action an untrusted input tried to AUTHORIZE (privilege escalation,
          authority pretext, third-party disclosure) or a product's permanent
          hard-stop. Refused outright; no in-band override.

The classification is by REVERSIBILITY and it lives in code — the deterministic
@sureel/approval-valve — never in a prompt. The snippet on the page is a
conceptual sketch of the shape, NOT a live endpoint or a real API call.

## The agent-discoverability wedge
Each product publishes a /reality.txt — "robots.txt told crawlers what they may
READ; reality.txt tells agents what they may DO." It makes the trust posture
legible to OTHER agents: an MCP client can read what a product will and won't
do before it ever calls. The verb is the wedge; the file is the handshake.
The reality.txt convention is proposed and openly published — NOT a standard,
NOT a certification, NOT a registry.

## The five (six) surfaces — the roll-call
- RadMail (radmail.ai) — won't send the wire. LIVE sandbox; email triage + MCP server.
- RadTask (radtask.ai) — won't delete the data. LIVE; an AI task cockpit.
- RadTalk (radtalk.ai) — won't wire by voice. COMING SOON; call-time firewall (code-backed).
- RadHealth (radhealth.ai) — won't prescribe. COMING SOON; clinic FRONT-OFFICE only.
- RadRobo (radro.ai) — won't pour the bleach. FOR-SHOW dream; restraint as a superpower.
- RadVeo (radveo.ai) — won't fake a face without consent. Family member; AI video/spokesperson.
(RadRank, radrank.ai, is also family — a neutral AI-answer scoreboard, LIVE.)

## The investor one-liner
"The first family of AI you trust BECAUSE it refuses." It rhymes with the
"own the rail, not the whole machine" doctrine: don't try to be every agent —
be the permission layer every agent rents. As agents proliferate, the scarce,
trusted thing isn't capability; it's the layer that can credibly say no.

## Honest status (read this before quoting)
- A DREAM IN PROGRESS. Some products are live sandboxes/previews; others are
  coming-soon or for-show. Pre-release framing throughout.
- A tool, not a guarantee. NEVER describe any Rad product as "fraud-proof",
  "guaranteed", "un-hackable", "HIPAA-certified", or "FedRAMP-authorized".
  Honest framing: "refuses the action classes that cause the losses."
- No adoption metrics, refusal counters, or activity numbers are claimed here;
  any such number anywhere in the family is hard-labeled real or illustrative.
- RadHealth is provider/clinic B2B only — no medical/efficacy claims, no patient
  language, no patient or outcome metrics, ever.
- The "try the classifier" widget on the page is a fixed deterministic lookup
  for illustration — no network call, no live model, no real API.

## Links
- https://radlife.ai/permission-as-a-service        (this category page)
- https://radlife.ai/permission-as-a-service/reality.txt
- https://radro.ai/the-line                           (The Amber Line verification)
- https://radlife.ai                                (the RadLife umbrella)
- Products: https://radmail.ai · https://radtask.ai · https://radtalk.ai ·
  https://radhealth.ai · https://radro.ai · https://radveo.ai · https://radrank.ai