# Authority Agent — Developer and Agent Resources

Machine-readable resources for the Authority Agent product surface operated by Capture the
Conversation LLC. If you are an AI agent or a developer integrating with Authority Agent,
this is the index.

## Product surfaces

| Surface | URL | What it is |
| --- | --- | --- |
| Category Codex sales site | https://codex.authorityagent.io | Public marketing site for the $2,500 one-week Category Codex engagement |
| Authority Engine | https://engine.authorityagent.io | MCP server that executes Authority Agent methodology server-side |
| Operator Console | https://console.authorityagent.io | Private operator console, access-gated |
| Company site | https://capturetheconversation.com | Capture the Conversation LLC |

## The Category Codex API

A public, read-only REST API over this site's content. No authentication, no rate limit, no
writes. Everything it returns is already published on the pages above — it is the same
information, structured for machines.

- **Base URL:** `https://codex.authorityagent.io`
- **OpenAPI 3.1 spec:** https://codex.authorityagent.io/openapi.json (also at `/api/openapi.json`)
- **Errors:** every failure returns JSON — `{ "error": { "code", "message", "hint", "docs", "status" } }` — never an HTML page
- **CORS:** open to all origins

| Operation | Method and path | What it does |
| --- | --- | --- |
| `getHealth` | `GET /api/health` | Liveness probe; returns the API version and where to find the spec, MCP endpoint, and llms.txt |
| `getOffer` | `GET /api/offer` | The Category Codex offer as structured data: price, duration, guarantee, deliverables, fit criteria, actions |
| `checkFit` | `POST /api/fit` | Applies the published fit criteria to a founder profile and returns a verdict with a per-criterion breakdown |
| `listPages` | `GET /api/pages` | Every public page with its HTML and Markdown URLs |
| `getPage` | `GET /api/pages/{slug}` | One page's metadata and full Markdown body |

Try it:

```bash
curl -s https://codex.authorityagent.io/api/offer
curl -s -X POST https://codex.authorityagent.io/api/fit \
  -H "Content-Type: application/json" \
  -d '{"annualRevenue":2400000,"founderLed":true,"role":"consultant"}'
```

`POST /api/fit` is a self-check, not a gate. It applies stated criteria mechanically and
says so in every response; no engagement is ever accepted or refused by an API.

## Category Codex MCP server

The same surface is available natively over the Model Context Protocol, so an agent can call
this site as tools instead of scraping it.

- **Endpoint:** `POST https://codex.authorityagent.io/mcp`
- **Transport:** Streamable HTTP, JSON-RPC 2.0
- **Manifest:** https://codex.authorityagent.io/.well-known/mcp.json
- **Authentication:** none. This server is public and read-only.
- **Protocol versions:** 2025-06-18, 2025-03-26, 2024-11-05

Tools: `get_offer`, `check_fit`, `list_pages`, `get_page`.

```bash
curl -s -X POST https://codex.authorityagent.io/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
```

This is a different server from the Authority Engine below. This one is public and describes
the offer; the Authority Engine is private, key-scoped, and runs client methodology.

## Authority Engine MCP server

The Authority Engine is a Model Context Protocol (MCP) server. Thin client skills — installed
in a client's own AI workspace — call the engine with a scoped API key; the methodology
executes server-side and only finished output is returned. Every response is watermarked per
client and usage-logged.

- **MCP endpoint:** `POST https://engine.authorityagent.io/mcp` (JSON-RPC 2.0)
- **Health check:** `GET https://engine.authorityagent.io/health` — returns `{"ok":true}` and
  the list of available method names, no authentication required
- **Authentication:** scoped API key issued per client, sent as a bearer token. Unauthenticated
  calls to `/mcp` return `401`. Keys are issued, rotated, and revoked by Capture the
  Conversation; there is no self-serve signup.
- **Transport:** HTTP POST, JSON-RPC 2.0, standard MCP `tools/list` and `tools/call` methods

### Available methods

`category_codex`, `jtbd_interview`, `voice_match`, `authority_voice_match`,
`bend_blend_break`, `ghostbase_draft`, `content_sawdust`, `support_triage`, `pr_booking`.

Each method loads its methodology server-side. Client context is supplied in the call;
methodology never leaves the engine.

### Requesting access

Engine access ships with the Install engagement. To request a key, email
**james@capturetheconversation.com** with the workspace you intend to connect. Access is not
self-serve and keys are not issued to anonymous requests.

## Agent-readable files on this site

| File | URL | Purpose |
| --- | --- | --- |
| llms.txt | https://codex.authorityagent.io/llms.txt | Index of this site for LLMs, including when to use it |
| llms-full.txt | https://codex.authorityagent.io/llms-full.txt | Full text of every public page in one file |
| agents.md | https://codex.authorityagent.io/agents.md | Agent instructions: when to reach for this site and how to call it |
| sitemap.xml | https://codex.authorityagent.io/sitemap.xml | XML sitemap of all indexable URLs |
| robots.txt | https://codex.authorityagent.io/robots.txt | Crawl policy and sitemap pointer |
| openapi.json | https://codex.authorityagent.io/openapi.json | OpenAPI 3.1 spec for the public API |
| mcp.json | https://codex.authorityagent.io/.well-known/mcp.json | MCP server manifest |

## Markdown content negotiation

Every public page on this site serves a Markdown variant from the same URL, per
[acceptmarkdown.com](https://acceptmarkdown.com):

```
curl -s -H "Accept: text/markdown" https://codex.authorityagent.io/
```

- Requests preferring `text/markdown` receive `Content-Type: text/markdown; charset=utf-8`
- Every negotiated response sends `Vary: Accept, Accept-Encoding`
- Accept `q`-values are honored; `text/html;q=1.0, text/markdown;q=0.5` returns HTML
- An `Accept` header that permits neither HTML nor Markdown returns `406 Not Acceptable`
- The `.md` twin of each page is also addressable directly: `/index.md`, `/about.md`,
  `/contact.md`, `/privacy.md`, `/developers.md`

Unknown paths return a real `404` with a short Markdown body pointing at the sitemap and
llms.txt, so a crawler that guesses wrong can recover in one hop.

## Structured data

Every page carries JSON-LD in `<script type="application/ld+json">`: an `Organization` node
for Capture the Conversation LLC with `contactPoint` and `address`, a `WebSite` node, and on
the homepage a `Service` node with `offers` for the $2,500 Category Codex engagement.