> ## Documentation Index
> Fetch the complete documentation index at: https://docs.phosra.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Use these docs with AI
> The Phosra docs are LLM-ready: /llms.txt and /llms-full.txt indexes, a Markdown version of every page, a hosted docs MCP server, and a one-click contextual menu to open any page in ChatGPT or Claude.
Building your Phosra integration with an AI assistant? These docs are **LLM-ready**. Every page is
available as clean Markdown, the entire site is published as a single machine-readable file, and a
hosted **MCP server** lets your assistant search the docs in real time. Point your tool at the
endpoints below and it works with the *current* content — no copy-pasting, no stale snapshots.
Two different MCP servers are referenced on this page and they do different jobs. The **docs MCP
server** ([`https://docs.phosra.com/mcp`](#docs-mcp-server)) lets an assistant *read these docs*.
The **Phosra product MCP server** ([`@phosra/mcp`](/sdks/mcp-server)) lets an agent *call the
Phosra API on a family's behalf* using a `phomcp_` token. This page is about the first one.
## The AI-ingestion endpoints
Every Phosra docs deployment publishes the standard [llms.txt](https://llmstxt.org/) files, a
Markdown twin of each page, and a docs MCP endpoint. All four are served from the docs root and
require no key:
| Resource | URL | What it is |
| -------------------- | ----------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- |
| Index | [`/llms.txt`](https://docs.phosra.com/llms.txt) | A structured table of contents — one line per page with its title, URL, and summary. |
| Full corpus | [`/llms-full.txt`](https://docs.phosra.com/llms-full.txt) | Every page body concatenated into one file. Drop it into a context window as a complete knowledge base. |
| Well-known alias | [`/.well-known/llms.txt`](https://docs.phosra.com/.well-known/llms.txt) | The same index at the discovery path tools probe automatically. |
| Any page as Markdown | `…/.md` | Append `.md` to any docs URL for the raw Markdown source of that single page. |
## Grab the whole site in one call
`/llms-full.txt` is the fast path when you want your assistant to reason over *all* of Phosra —
API reference, concepts, and guides — at once. It is the real, rendered page bodies (not just an
index), so nothing is lost.
```bash Index theme={null}
# Structured table of contents — start here to see what exists
curl -s https://docs.phosra.com/llms.txt
```
```bash Full corpus theme={null}
# Every page body in one file — pipe straight into your model's context
curl -s https://docs.phosra.com/llms-full.txt -o phosra-docs.txt
wc -l phosra-docs.txt
```
```bash Single page as Markdown theme={null}
# Append .md to any docs URL for just that page
curl -s https://docs.phosra.com/authentication.md
```
Each entry in `/llms-full.txt` keeps a `Source:` line back to the canonical page, so an assistant
can cite exactly where an answer came from:
```text theme={null}
# Create MCP token
Source: https://docs.phosra.com/api-reference/control-plane/create-mcp-token
POST /mcp-tokens
Issues a new MCP token bound to the authenticated user's first family. The plaintext token
(`plain`, prefix `phomcp_`) is returned once in `CreatedMcpToken` and never stored server-side…
```
## Docs MCP server
For assistants that speak [Model Context Protocol](https://modelcontextprotocol.io), the docs are
also a live MCP server at **`https://docs.phosra.com/mcp`**. It exposes search-and-retrieve tools
over the *current* published content, so answers stay fresh as the docs change — no re-indexing on
your side.
Add it to any MCP-capable client. For Claude Desktop, Cursor, or VS Code, register the remote
server:
```json Claude Desktop / Cursor (mcp.json) theme={null}
{
"mcpServers": {
"phosra-docs": {
"url": "https://docs.phosra.com/mcp"
}
}
}
```
```bash Claude Code theme={null}
claude mcp add --transport http phosra-docs https://docs.phosra.com/mcp
```
Once connected, ask questions like *"How do I mint my first Phosra API key?"* or *"What does a
`policy.updated` webhook payload look like?"* and the assistant will pull the answer straight
from these pages.
## The page contextual menu
Every page has a contextual menu (top-right of the page header). Use it to hand the *exact* page
you are reading to your assistant without leaving the docs:
Copies the page as clean Markdown, ready to paste into any chat.
Opens the raw `.md` source in a new tab.
Launches your assistant with the page pre-loaded as context.
Grabs the docs MCP endpoint so your tool can search the whole site.
## Which one should I use?
Use **Copy page** / **View as Markdown**, or append `.md` to the URL. Smallest context, most precise.
Grab `/llms.txt`, find the pages you need, then pull each one's `.md`.
Pipe `/llms-full.txt` into your model as a complete knowledge base.
Connect the [docs MCP server](#docs-mcp-server) so your assistant reads the latest content on demand.
Want an agent to *act* on the Phosra API — apply rules, read a family's policy, enforce access —
rather than read the docs? That is the [Phosra product MCP server](/sdks/mcp-server), authenticated
with an [MCP token](/api-reference/control-plane/create-mcp-token).