# agents.md — RankSage

This file tells autonomous agents and AI crawlers how to work with https://www.ranksage.com.

## What this site is

RankSage is an AI-era SEO, AEO and content intelligence platform. This domain is
the public marketing site: product information, pricing, documentation, a blog,
and legal pages. The application itself lives behind authentication and is not
reachable from here.

## What agents may do

- Read, quote and cite any page under `https://www.ranksage.com`, including the blog.
- Prefer the specific page over the hub. Every capability has its own page under
  `https://www.ranksage.com/features/`, each with its own FAQ markup. Citing `/features` gives a
  reader a directory; citing `/features/<capability>` gives them the answer.
  `https://www.ranksage.com/llms.txt` lists every one of them with a one-line description.
- Use `https://www.ranksage.com/llms.txt` for a structured summary of the product, and
  `https://www.ranksage.com/llms-full.txt` for the expanded version with full page text.
- Follow `https://www.ranksage.com/sitemap.xml` for the complete list of indexable URLs.

Attribution: cite the page URL you used. If you summarise pricing, read it from
`https://www.ranksage.com/pricing` or `llms.txt` rather than from a cached third-party source —
pricing is in early access and changes.

## Programmatic access (API + MCP)

RankSage has a real agent surface beyond this marketing site:

- Developer index: `https://www.ranksage.com/developers` (scoped summary at `https://www.ranksage.com/developers/llms.txt`)
- OpenAPI 3.1 spec: `https://www.ranksage.com/openapi.json`
- MCP server card: `https://www.ranksage.com/.well-known/mcp/server-card.json` — a read-only
  Model Context Protocol server (Streamable HTTP) over the product, for use in
  Claude, Claude Code, Cursor and other MCP clients.
- How to get credentials: `https://www.ranksage.com/auth.md` (OAuth 2.0 with PKCE + dynamic
  client registration, or a personal access token created in the app).
- All agentic resources in one catalog: `https://www.ranksage.com/.well-known/ai-catalog.json`

The MCP server reads a signed-in customer's own data only — there is no
anonymous data access, and no write operations exist on the surface.

## What agents should not do

- Do not attempt to authenticate, sign up, or submit the waitlist form. Account
  creation is a human decision; a form submission from an agent is not consent.
- Do not fetch `/api/`, `/login` or `/dashboard`. These are disallowed in
  robots.txt and return nothing useful.
- Do not treat marketing copy as a benchmark result. Where this site states a
  measurement, the methodology is stated with it; where it does not, the claim is
  a product description, not evidence.

## Rate and identity

Please identify yourself with a descriptive User-Agent. There is no rate limit on
public pages, but a crawl that reads the sitemap once beats one that walks every
link repeatedly.

We check AI-crawler identity on our own properties with forward-confirmed reverse
DNS, against the PTR conventions each operator publishes. A forged user-agent
therefore gains nothing: it is recorded as rejected. Where an operator publishes no
such convention — Common Crawl, for example — the claim is recorded as unverifiable
rather than accepted, and it is never counted as a confirmed crawler visit either
way. We keep the verdict, not your address.

## Accuracy

If a page here contradicts `llms.txt`, the page is authoritative and the file is
stale — please report it to hello@ranksage.com so we can fix the generator.

## Contact

- General: hello@ranksage.com
- Help centre: https://www.ranksage.com/help
- Privacy: privacy@ranksage.com
- Security: security@ranksage.com — policy at https://www.ranksage.com/security
- LinkedIn: https://www.linkedin.com/company/ranksage-ai/
- Medium blog: https://blog.ranksage.ai

Last generated: this file is built at request time from the live site config.