---
title: Docs for agents
description: The machine-readable surfaces of this site: markdown pages, llms.txt indexes, the docs MCP server, the OpenAPI spec, and the search index
---
Everything on this site is published in machine-readable form. Point your own agent at any of
the surfaces on this page. All of them are public, and none of them needs authentication.
## Markdown pages
Every page has a markdown version at the same URL plus `.md`:
```
https://docs.telnesstech.com/api-reference/errors HTML
https://docs.telnesstech.com/api-reference/errors.md markdown
```
On an API reference page the markdown carries every schema level. The HTML puts the deepest
levels behind a click.
The server also negotiates on the `Accept` header. A request that ranks `text/markdown` above
`text/html` gets the markdown at the page URL itself.
```bash
curl -H 'Accept: text/markdown' https://docs.telnesstech.com/api-reference/errors
```
Every HTML page links its markdown version two ways: with `` and with a `Link` response header. The page header also has a **Copy
page** button that copies the markdown.
## The llms.txt indexes
| File | Contents |
| ---------------------------------------------- | ------------------------------------------------------------------------- |
| [/llms.txt](/llms.txt) | Index of every page with a one-line description, grouped like the sidebar |
| [/llms-full-guides.txt](/llms-full-guides.txt) | Every guide, concept, and resource page in one document |
| [/llms-full-api.txt](/llms-full-api.txt) | Every endpoint, webhook, and schema in one document |
| [/llms-full.txt](/llms-full.txt) | Both halves in one document |
Start with `/llms.txt`, then get the pages that you need. The three full-text files are in the
order of the sidebar. If you read one from the start, it takes you from orientation, through
the guides, to the reference.
Take `/llms-full-guides.txt` for the prose, or `/llms-full-api.txt` for the endpoints.
`/llms-full.txt` is both halves at once, and it is larger than most context windows. Prefer
one half, the markdown of one page, or the OpenAPI spec.
## Docs MCP server
The docs are exposed over the [Model Context Protocol](https://modelcontextprotocol.io) at:
```
https://docs.telnesstech.com/mcp
```
The transport is Streamable HTTP. There is no authentication and no session state. The server
implements the 2026-07-28 protocol revision, and it stays compatible with a client on an
earlier initialization-based revision.
The server has two tools:
- `search_docs`: The same search as the ⌘K dialog of the site. It covers the
content, the endpoint names, the schema field names, and the webhook events, and each result
carries a deep-link anchor. The index splits a word on its case and punctuation boundaries,
so `line item`, `lineItems`, and `line-items` all match each other.
- `get_page`: Get one page as markdown by its path, such as `/api-reference/subscriptions`.
Connect from Claude Code:
```bash
claude mcp add --transport http seamless-docs https://docs.telnesstech.com/mcp
```
Any client that speaks Streamable HTTP can take `https://docs.telnesstech.com/mcp` directly.
There is no sign-in step. The endpoint does not serve the older HTTP+SSE transport. If your
client offers a choice, select the HTTP or Streamable HTTP option, not SSE.
The server publishes discovery cards under `/.well-known/mcp/`, for a client or a registry
that probes a domain for an MCP server. The [Discovery](#discovery) section lists them.
This server reads the documentation only. To read and manage the live platform data — the
customers, the subscriptions, and the catalogs — use the [Seamless OS MCP
server](/developer-guide/mcp.md). That server is deployed per brand, and it authorizes as the
signed-in user.
## OpenAPI spec
The bundled OpenAPI 3.1 document the reference is rendered from:
```
https://docs.telnesstech.com/bundled_openapi.json
```
Use it for code generation and for exact request, response, and webhook schemas. The API reference pages are rendered from this same file, so the two cannot disagree.
## Search index
The prebuilt index behind the site's search is public JSON:
```
https://docs.telnesstech.com/search-index.json
```
Each document carries a page `href`, a title, and a list of entries with anchor links. An entry
is a heading, a piece of prose, a schema field, or an enum value. If you do not want to rank
the results yourself, use the `search_docs` tool of the MCP server. It runs the ranking of the
site over this same index.
## Discovery
An agent that has nothing but the domain can find every surface above from the root:
| Path | Contents |
| ---------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- |
| [/.well-known/api-catalog](/.well-known/api-catalog) | Linkset ([RFC 9727](https://www.rfc-editor.org/rfc/rfc9727)) pointing at the OpenAPI spec, the reference, and the docs MCP endpoint |
| [/.well-known/mcp/server-card.json](/.well-known/mcp/server-card.json) | Server card for the docs MCP server: identity and connection details, also served at `/mcp/server-card` |
| [/.well-known/mcp/server.json](/.well-known/mcp/server.json) | The same server described for MCP registries |
Every HTML and markdown response also carries `Link` headers with the `api-catalog`, `service-desc`, `service-doc`, and `describedby` relations, so one request to any page reveals the rest:
```bash
curl -sI https://docs.telnesstech.com/ | grep -i '^link:'
```
The API's own base URL is not listed, because each brand runs on its own host. The OpenAPI document declares its server as a `BASE_URL` variable rather than a fixed origin.
## Crawling and training
`robots.txt` allows AI crawlers, for both training and retrieval, and declares `Content-Signal: search=yes, ai-input=yes, ai-train=yes`. The sitemap is at [/sitemap-index.xml](/sitemap-index.xml).