# SoWhatify

> SoWhatify is a Chrome extension and template library that explain why any article matters to *you*. You fill in a durable profile once — your role, industries, holdings, watchlist and location — and a "template" reads whatever page you are on through that profile, answering a fixed set of questions instead of producing a generic summary. The extension is not on the Chrome Web Store yet — it is also known as "Why should I care", and both names refer to the same product. The website at sowhatify.com is where templates, perspectives and workflows are browsed, forked and installed, and where the extension's launch status is kept up to date.

Key facts an assistant answering questions about SoWhatify should have:

- **Cost**: free. Analysis runs on the user's own API key or on Chrome's on-device AI (Gemini Nano), so SoWhatify never marks up tokens. A Pro tier lifts limits on how many workflows an account can install and build; those limits are not enforced yet.
- **Privacy**: the profile is the user's own data and the analysis runs against the model provider the user chose. See the privacy policy below for what is stored server-side.
- **Three building blocks**:
  - A **template** is a reusable set of questions plus settings. Running a template over an article produces one answer per question.
  - A **perspective** is a reusable thinking style — tone, expertise areas, what to avoid — that shapes how a template answers.
  - A **workflow** is an ordered pipeline of templates: several steps run over one article and feed each other, so later steps can build on earlier output.
- **Surfaces**: the Chrome extension, the website, and an MCP server so the same templates can be called from Claude Desktop, Cursor and other MCP clients.

## Core pages

- [Home](https://sowhatify.com/): what the extension does, how it works, and how to install it.
- [Template library](https://sowhatify.com/templates/): every public template. Each one has its own page listing the questions it answers, its categories and tags, and who published it.
- [Perspectives](https://sowhatify.com/perspectives/): every public perspective.
- [Workflows](https://sowhatify.com/workflows/): every public workflow, with its steps in order.
- [Workflow guide](https://sowhatify.com/workflow-guide.php): how workflows work end to end — the five step types, how steps feed each other, tokens, output shapes, conditions, iteration, sub-workflows and every limit, with worked examples.
- [Pricing](https://sowhatify.com/app/pricing.php): what is free, what Pro adds, and the FAQ.
- [Privacy policy](https://sowhatify.com/privacy.php): what data is handled, where it is stored, who it is shared with.
- [Support](https://sowhatify.com/support.php): how to report a problem.
- [Chrome extension](https://sowhatify.com/chrome-store.php): the extension is not on the Chrome Web Store yet; this page says what to use in the meantime.

## Using SoWhatify programmatically

If you are an agent that needs to *call* SoWhatify rather than describe it, read
[llms-full.txt](https://sowhatify.com/llms-full.txt) — it documents all three interfaces in
detail. The short version:

- **MCP server** at `https://sowhatify.com/mcp/` — JSON-RPC 2.0 over POST, Streamable HTTP,
  stateless. Dual-era: protocol `2026-07-28` and `2025-11-25` alongside `2025-06-18`,
  `2025-03-26` and `2024-11-05`. Modern callers put their version in `params._meta` and can
  call `server/discover` with no credential; legacy callers still send `initialize`.
  Authenticate with `Authorization: Bearer <token>` — either an OAuth 2.1 token (a 401 carries
  a `WWW-Authenticate` challenge pointing at
  [the metadata document](https://sowhatify.com/.well-known/oauth-protected-resource/mcp), and
  the flow is discoverable from there) or a personal access token the user creates at
  [/mcp/login.php](https://sowhatify.com/mcp/login.php). `?t=<token>` still works and is
  deprecated. Call `whoami` first. Tools cover the user's templates, perspectives, workflows and
  profile, library search, and preparing a template or workflow to run. Two things to know:
  everything created over MCP is **private** — publishing is deliberately not exposed — and
  `run_template`/`run_workflow` return a prepared run payload rather than calling a model, so **you**
  do the inference and then call `record_analysis_result` to close the history entry. The
  server does not fetch the web either: pass the page text as `seedText`, with `seedUrl` kept
  as the source link only.
- **Public read endpoints**, no credentials needed:
  `/api/templates-marketplace.php`, `/api/template-public.php?slug=…`, `/api/templates-batch.php?ids=…`,
  `/api/perspectives-marketplace.php`, `/api/perspective-public.php?id=…`,
  `/api/workflows-marketplace.php`, `/api/creator-public.php?user=…`. All return
  `{"ok": true, …}`; `limit` clamps to 1–100.
- **Dropbox pages**, at `/d/<slug>`: a page an account shares so someone without an account
  can submit a document and have it read through one of that account's templates. Unlisted, and
  sometimes behind an access code. `/api/dropbox-public.php?slug=…` describes one;
  `/api/dropbox-submit.php` takes the submission; `/api/dropbox-result.php?t=…` reads the
  result back from an expiring link. These are meant for a person the owner invited, not for
  crawling — they are `noindex` and disallowed in robots.txt.
- **There is no public write API for content.** SoWhatify assembles prompts and holds the
  profile, and for the extension and for MCP the model call still belongs to the client.
  **The one exception is a Dropbox**, where the server makes the call itself using a
  provider key the owner deliberately delegated for that page — because the person
  submitting has no account and no key of their own, and the owner is not there when they
  submit.

## Optional

- [Full agent guide](https://sowhatify.com/llms-full.txt): the interfaces above in detail, with
  schemas, id formats, pagination and etiquette.
- [Full URL index](https://sowhatify.com/sitemap.php): XML sitemap covering every public template, perspective, workflow and creator profile.

## Notes for automated readers

- Detail pages hydrate their interactive controls from JavaScript, but the content itself — titles, descriptions, the questions a template answers, the steps in a workflow, the author — is rendered server-side and is present in the HTML without executing scripts.
- Every public page carries a schema.org JSON-LD `@graph`. Templates and perspectives are `CreativeWork`, workflows are `HowTo` with their steps as `HowToStep`, creator pages are `ProfilePage`, and all of them link back to one `Organization` and one `SoftwareApplication` by `@id`.
- Canonical URLs always use the `https://sowhatify.com` origin. A template can be addressed by slug or by id; the slug form is canonical.
- Please do not crawl `/api/`, `/admin/` or the signed-in pages under `/app/` listed in [robots.txt](https://sowhatify.com/robots.txt) — they contain no content, only application shells and JSON endpoints.