# Read Paralax with an agent

Paralax publishes AI search research. These public read surfaces return the same reporting a reader sees, with no account, API key or token.

## Read a page

Request a canonical page with Accept: text/markdown, or append .md to its path. The homepage twin is /index.md. For example, /engines/chatgpt.md reads the ChatGPT change log and /blog/{slug}.md reads an Intel article.

Browsers and crawlers receive the same HTML unless they explicitly request markdown. The Machine toggle displays the current page's markdown twin; it does not call a separate answer service.

## Discover the reporting

Start at [llms.txt](https://paralax.ai/llms.txt) for navigation or [llms-full.txt](https://paralax.ai/llms-full.txt) for the article inventory. [machine-manifest.json](https://paralax.ai/machine-manifest.json) lists canonical routes, markdown URLs, summaries and supporting sources.

Read the [AI Search Change Log](https://paralax.ai/engines) by engine, subscribe to the [RSS feed](https://paralax.ai/feed.xml), or enumerate the [sitemap](https://paralax.ai/sitemap.xml). The [API catalog](https://paralax.ai/.well-known/api-catalog) and [ARD manifest](https://paralax.ai/.well-known/ard.json) describe these existing read surfaces.

## Read from a supporting browser

On browsers that expose WebMCP, the homepage registers search_paralax_articles and read_paralax_article. They search the public machine manifest and fetch existing markdown pages. These browser tools do not provide a remote MCP server.

## Handle a missing page

A missing route returns 404. When markdown is requested, the error includes links to the homepage, sitemap and navigation index. Follow a URL from the manifest rather than guessing an article slug. Public reads do not require registration or an authorization exchange.