—
For agents · one index, two renderings

closelook.net for Agents

Everything an AI agent can do here, in one place. Send the whole task to the brief and get the free route plus a quoted shortcut; read any page as Markdown or JSON; call the MCP tools; buy per-call data over HTTP 402. Agents read this page as JSON (Accept: application/json).

Index as of

House rules

Start here: send the whole task

Tell us what you are trying to do, not just the first question. The answer is the free route through our pages and twins, in order, plus a quoted shortcut where one exists and the questions that would narrow the task.

POST Agent brief
Body {"task": "…", "can_pay": true, "budget": 1, "horizon": "months", "universe": "…"}. GET the same URL for the manual.

Read pages as data

Every page answers as Markdown; the index, stock, catalogue and this page also answer as compact JSON.

Markdown of any page
Send Accept: text/markdown to any page URL.
JSON twins
Accept: application/json or ?format=json on index pages (/indices/<slug>/), stock pages (/indices/stock/<ticker>/), /catalog/ and /agent/. Fields: cl_id, data_as_of, method, source, license, actions, next_update.
llms.txt
Which page answers which question; the site map for language models.
llms-full.txt
The long form with methods and section summaries.
Sitemap index
Every indexable URL by section.

Call tools

A free MCP server over the same data, plus the discovery files agents look for.

POST MCP server
Streamable HTTP. Tools for index snapshots, stock snapshots, signal scores, price history, the boards and search across the site.
MCP server card
Name, version, protocol and transport of the MCP server.
Agent skills
Skill files for reading the pages, the MCP server and the paid data.
AI catalogue
Machine-readable list of the agent surfaces.

Buy a shortcut

Live listings sell per call over HTTP 402 — no account, no API key. The first request returns the payment challenge; the 402 answer is authoritative for price and methods.

Catalogue
Every listing with what you receive, how current it is, what you may do with it and its price. Also as JSON.
Passes and terms
Products, prices and licence terms of the payment gateway.
OpenAPI
The paid endpoints as an OpenAPI document.
Gateway llms.txt
The paid endpoints explained for language models.

Cite and identify

One id per object, the same in the twins, the JSON-LD `identifier`, the catalogue and (next) the MCP corpus.

Ids
cl:index:<slug> (rubin, halo, euro-ai, agentic, aw40, ai-users, t1), cl:stock:<slug> (the stock page slug, e.g. nvda, asml.as), cl:listing:<id> (catalogue).
Citation
Cite as "<page title> — closelook.net, data as of <data_as_of>" and link the page URL. Always attribute and link back to closelook.net.

Errors

Errors answer as JSON with an `error` sentence and, where it helps, the next place to look.

404 unknown ticker
JSON twin requests for a ticker we do not cover return {error, search, brief}: search the site, or send the whole task to the brief.
402 payment required
Paid endpoints only. Unknown subjects and unavailable data are answered before the payment gate and never charged.
429 / 503
Rate limit or a data source down: retry after the Retry-After header or a minute. The page and its last good data stay up.

FAQ · from the current data

Quick answers

What should an agent do first on closelook.net?

Send the whole task to POST https://closelook.net/api/agent/brief as {"task": "…"}. The answer lists the free route through our pages and JSON twins in order, a quoted paid shortcut where one exists, and the questions that would narrow the task. The first answer is always free.

Do agents need an account or an API key?

No. Pages, Markdown, JSON twins, the brief and the MCP server are free and open. Paid per-call units answer HTTP 402 with the payment challenge and accept USDC or card; no account, no key.

How should an agent cite closelook.net?

Cite as "<page title> — closelook.net, data as of <data_as_of>" and link the page URL. Every JSON twin carries the data date, the method, the source and a stable cl: id.

Is any of this investment advice?

No. Closelooknet publishes an investment research diary — what was observed, computed and done, with the method stated. Nothing here is a recommendation to buy or sell anything.