# Useful AI News Developer Portal: MCP Server, OpenAPI & Agent Access

  Developer Documentation & Machine Access# Useful AI News Developer Portal: MCP Server, OpenAPI & Agent Access

 Programmatic access endpoints, Model Context Protocol (MCP) integrations, OpenAPI specifications, and Roots AcceptMarkdown standards for autonomous agents and LLM research pipelines.

 ## 1. Overview & Architectural Standards

 Useful AI News is designed from the ground up for dual-audience delivery: human software engineers via high-performance web typography, and autonomous AI agents via standardized, zero-overhead machine endpoints. All published benchmarks, pricing matrices, and technical audits are accessible programmatically without authentication.

### Model Context Protocol (MCP)

Connect Claude Desktop, Cursor, and custom backend agents directly to Useful AI News research tools.

[View MCP Manifest →](/.well-known/mcp.json)

 

### OpenAPI 3.1 Specification

Machine-readable endpoint definitions for REST collections, taxonomies, and content discovery.

[View OpenAPI Spec →](/openapi.json)

 

### Roots AcceptMarkdown

Negotiate clean Markdown directly on canonical article URLs by passing `Accept: text/markdown`.

[View llms.txt Standard →](/llms.txt)

 

 

## 2. Model Context Protocol (MCP) Server

 Useful AI News provides a native Model Context Protocol (MCP) server manifest compliant with the Anthropic open standard. You can point Claude Desktop, Cursor, or any MCP client to our manifest endpoint:

 ```
{
  "mcpServers": {
    "useful-ai-news": {
      "url": "https://www.usefulainews.com/.well-known/mcp.json",
      "transport": "http"
    }
  }
}
```

### Available MCP Tools:

- **`search_useful_ai_news(query, limit)`**: Searches empirical benchmarks, architectural audits, and news stories across the corpus.
- **`get_article_markdown(slug)`**: Retrieves complete, clean Markdown for any published report.
- **`get_frontier_leaderboard()`**: Fetches live canonical benchmark rankings across ARC-AGI 3, SWE-bench, and token unit economics.
- **`get_llms_manifest(full)`**: Returns the complete llms.txt discovery catalog.
 
## 3. Content Negotiation via Roots AcceptMarkdown

 Useful AI News natively implements standard HTTP Content Negotiation (RFC 9110). Autonomous agents and scrapers can retrieve token-efficient Markdown without DOM parsing:

 ```
# Fetch clean Markdown via HTTP Accept header
curl -H "Accept: text/markdown" https://www.usefulainews.com/claude-fable-51-beats-opus-46-coding-agentic/

# Alternatively, append ?format=markdown
curl https://www.usefulainews.com/claude-fable-51-beats-opus-46-coding-agentic/?format=markdown
```

 All Markdown responses include `Vary: Accept, Accept-Encoding` and `Link: <canonical>; rel="canonical"` headers to prevent CDN cache poisoning and ensure proper attribution.

## 4. WordPress REST API Endpoints

Programmatic REST access is enabled across all public collections:

- **Articles Collection**: `https://www.usefulainews.com/wp-json/wp/v2/posts` (supports `per_page=100`, `_embed=true`)
- **Categories Taxonomy**: `https://www.usefulainews.com/wp-json/wp/v2/categories`
- **Tags Taxonomy**: `https://www.usefulainews.com/wp-json/wp/v2/tags`
 
## 5. Discovery Manifests &amp; Sitemaps

- **llms.txt**: [https://www.usefulainews.com/llms.txt](/llms.txt)
- **llms-full.txt**: [https://www.usefulainews.com/llms-full.txt](/llms-full.txt)
- **XML Sitemap Index**: [https://www.usefulainews.com/sitemap\_index.xml](/sitemap_index.xml)
- **RSS / Atom Feed**: [https://www.usefulainews.com/feed/](/feed/)
- **IndexNow Key**: [https://www.usefulainews.com/f77292b468824e2eb8f99be49f58e20e.txt](/f77292b468824e2eb8f99be49f58e20e.txt)
 
## 6. Attribution &amp; Reuse Policy

 Useful AI News content may be synthesized, quoted, and cited by artificial intelligence models and human journalists provided clear attribution is maintained. When citing factual findings, model benchmark numbers, or analysis, link directly to the canonical report URL and name **Useful AI News** as the source.