# llms.txt, Markdown and discovery files

> This documentation as plain text and as Markdown pages for retrieval and system prompts, the machine-readable API contract, and the files that let an agent find Chord's tools from the domain alone.

Source: https://getchord.xyz/docs/agents/llms-txt/

| File | What | Size |
| --- | --- | --- |
| [`/llms.txt`](/llms.txt) (also [`/docs/llms.txt`](/docs/llms.txt)) | What Chord is, then every page of this documentation with a one-line description, the API, the agent integrations and the app ([llmstxt.org](https://llmstxt.org) format) | small |
| `/docs/<page>.md` | Any page as Markdown: add `.md` to the page's path (`/docs/agents/mcp/` → [`/docs/agents/mcp.md`](/docs/agents/mcp.md)), or ask for the page with `Accept: text/markdown`. API operations too ([`/docs/api-reference.md`](/docs/api-reference.md) lists them) | per page |
| [`/llms-full.txt`](/llms-full.txt) (also `/docs/llms-full.txt`) | Every page of this documentation, concatenated | large |
| [`/docs/llms-small.txt`](/docs/llms-small.txt) | The same, minified — for a context window that needs the whole thing | medium |
| [`/openapi.json`](/openapi.json) (also `/docs/openapi.json`) | The OpenAPI 3.1 contract: every endpoint, parameter, schema and example | large |

## Discovery files

An agent that only knows `getchord.xyz` finds everything else from these:

| File | Standard | Points to |
| --- | --- | --- |
| [`/.well-known/api-catalog`](/.well-known/api-catalog) | RFC 9727 (linkset) | the REST API with its OpenAPI document, docs and status; both MCP servers; the skill index |
| [`/.well-known/ai-catalog.json`](/.well-known/ai-catalog.json) | AI Catalog ([MCP SEP-2127](https://modelcontextprotocol.io/seps/2127-mcp-server-cards)) | the server cards of the [MCP server](/docs/agents/mcp/) and the [Docs MCP](/docs/agents/docs-mcp/) |
| [`/.well-known/agent-skills/index.json`](/.well-known/agent-skills/index.json) | [Agent Skills discovery](https://github.com/cloudflare/agent-skills-discovery-rfc) | the [Chord skill](/docs/agents/skill/) archive and its SHA-256 digest |
| [`/robots.txt`](/robots.txt) | robots + Content Signals | crawlers, assistants and agents allowed (`search`, `ai-input`, `ai-train` = yes); both sitemaps |

`api.getchord.xyz` answers `/.well-known/api-catalog`, `/.well-known/ai-catalog.json` and `/robots.txt` too, and `GET https://api.getchord.xyz/` returns the same links as JSON.

## In a system prompt

Paste `llms.txt` and let the model fetch the `.md` pages it needs, or paste `llms-small.txt` when the model must answer without fetching. Add the two facts models get wrong most:

```
Chord API base: https://api.getchord.xyz/v1. Amounts are decimal strings in the token's smallest unit.
On Arc (chain 5042) USDC is the gas token; send the ERC-20 predeploy 0x3600000000000000000000000000000000000000 (6 decimals); the zero address is an alias for it and transactions always have value 0.
```

## For codegen by a model

Give it `openapi.json`. Every operation has an `operationId`, constraints on parameters (enums, ranges, patterns), a response schema, and a real example captured from the live API.

## For a retrieval index

Fetch `llms-full.txt` and split on the page separators; each page starts with its title and URL. Refresh daily — the file is rebuilt with every deploy of the docs.