For agents

Agent-friendly documentation

How an AI agent or script can read these docs as Markdown, discover every page, and call the API.

These docs are built to be read by programs as well as people. Nothing here needs a special client.

Markdown for every page#

Append .md to any page address:

Page	Markdown
`https://docs.aggre.org/quickstart`	`https://docs.aggre.org/quickstart.md`
`https://docs.aggre.org/api/data/market`	`https://docs.aggre.org/api/data/market.md`
`https://docs.aggre.org/`	`https://docs.aggre.org/index.md`

The Markdown version has the title as # heading, a one-line summary, the page's canonical address, then the body. Links inside it are absolute and point to other .md pages, so an agent can follow them without leaving Markdown. Code blocks are fenced with their language. The response is text/markdown; charset=utf-8, with a Link: <page>; rel="canonical" header and X-Robots-Tag: noindex so search engines index the HTML.

You can also ask for Markdown at the page's own address with an Accept header:

bash

curl -s -H "Accept: text/markdown" https://docs.aggre.org/concepts/errors

Every HTML page advertises its twin with <link rel="alternate" type="text/markdown" href="...">, and has a Copy page button that puts the Markdown on the clipboard.

Discovering pages#

URL	What
`/llms.txt`	An index in the llms.txt format: every page with its description and `.md` link, grouped by section. Start here.
`/llms-full.txt`	The complete text of every page in one file, for loading into a context window or an index.
`/sitemap.xml`	The HTML pages.
`/search-index.json`	The search index: one entry per section with a short excerpt.

Machine-readable API description#

/openapi/data.yaml (and .json): every route of the data API with parameters, and $refs to the row schemas.
/openapi/accounts.yaml: the console's control-plane routes.
/schemas/{name}.json: the JSON Schema of every row.

The reference pages are generated from these files, so the prose and the files cannot drift apart.

Calling the API from an agent#

Give the agent an API key through its tool configuration or environment, never in a prompt or a repository (Authentication).
Start from GET /v1/datasets to see what exists, and read meta.status on every answer.
Follow the rules in Best practices for agents: strict parameters, credit shown with data, terms_unread is not retryable.
For legal retrieval and judgments there is also an MCP-style tool interface: MCP compatibility.

Source: content/agents/overview.md