v0 · automotive vertical

The car-news MCP server for AI agents.

Install once. Your agent answers automotive questions with cited primary sources, deduped events, and a source-quote per fact — in one tool call.

Install

Claude Code

claude mcp add mcpnewsroom https://mcpnewsroom.com/mcp

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "mcpnewsroom": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcpnewsroom.com/mcp"]
    }
  }
}

What your agent gets

A single tool, search_events(query, limit, since)
Structured event records — title, 50-word brief, 300-word summary, key facts, importance, confidence
Source-quote per key fact — continuous, unedited, verbatim from the source body
Cross-source deduplication — one event, all citations attached
Confidence stamped post-synthesis, rule-based by source class (not LLM-generated)
Recency × importance ranking with 30-day half-life
Streamable HTTP MCP transport. Free. No API key.

One example

Agent calls search_events("BMW recalls", since="1095d"). Response (abridged):

{
  "query_interpretation": "Most recent BMW recall events ranked by
                           recency × importance over last 1095d.",
  "primary_event": {
    "title": "BMW recalls 2022–2025 i4, iX, i7, i5 for motor software defect",
    "event_type": "recall",
    "importance": 0.68,
    "confidence": 1.0,
    "sources": [{
      "publisher": "NHTSA",
      "type": "primary",
      "url": "https://www.nhtsa.gov/recalls?nhtsaId=25V395000"
    }],
    "key_facts": [{
      "fact": "Recall affects 2022–2025 i4, iX, i7, and i5 models",
      "source_quote": "BMW of North America, LLC is recalling certain
                       2022-2025 i4, iX, i7, and i5 vehicles...",
      "entailment": "direct"
    }, ...]
  },
  "related_events": [...]
}

Sources at v0

Source	Class	Confidence
BMW Group PressClub	Primary OEM	1.00
NHTSA Recalls API	Primary regulator	1.00
Electrek	Specialty EV	0.75
InsideEVs	Specialty EV	0.75

Coverage expands monthly. Tesla.com is bot-walled at the source — Tesla coverage flows through Electrek and InsideEVs at 0.75 confidence.

How it works

1 Ingest — fetch from primary sources on agent query (lazy). Raw HTML archived to R2 immediately.

2 Dedup — URL-hash check first, then semantic vector search at >0.85 cosine similarity over the last 30 days. Match → merge sources into existing event.

3 Synthesize — Haiku 4.5 reads the body and produces JSON. Hard rule: every key fact must include a continuous, unedited source-quote.

4 Stamp — confidence assigned post-synthesis by source class (1.0 primary, 0.75 specialty, …). Importance ranked per editorial rubric.

Who it's for

Developers building AI agents that field automotive questions — Claude Desktop, Cursor, Cline, Claude Code, and custom agent stacks. Free for any usage; data hygiene rules apply (we log only query terms, never full prompt content).