NHTSA Vehicle Safety Intelligence MCP Server

An Apify actor that exposes the National Highway Traffic Safety Administration (NHTSA) public API as a Model Context Protocol (MCP) server. AI assistants, automation pipelines, and developer tools can call four structured tools to retrieve vehicle recalls, consumer safety complaints, 5-star crash ratings, and component-based recall searches — all in real time, with zero API keys required.

---

What This Actor Does

The NHTSA maintains one of the most important public safety databases in the United States. Every manufacturer-issued vehicle recall, every consumer complaint submitted to the government, and every crash test rating produced in NHTSA's labs is available through their public API. This actor wraps that API in an MCP-compliant HTTP server so that Claude, GPT-4, LangChain agents, or any MCP-compatible client can call it directly with structured tool calls.

No API key. No registration. No rate-limit tokens to manage. The NHTSA API is fully open.

---

MCP Tools Available

Point your MCP client at http://<actor-run-url>:4321 after setting serveMcp: true.

1. get_recalls

Retrieve all NHTSA safety recalls for a given vehicle.

Parameters:

Name	Type	Required	Description
make	string	Yes	Manufacturer name (e.g. Toyota, Ford, Tesla)
model	string	No	Model name (e.g. Camry, F-150)
model_year	string	No	Four-digit year (e.g. 2022)
limit	integer	No	Max results (default 50, max 500)

Returns: List of recall records with fields: recall_id, manufacturer, subject, component, summary, consequence, remedy, report_date, units_affected, make, model, model_year, source.

Example call:

{
  "name": "get_recalls",
  "arguments": {
    "make": "Tesla",
    "model": "Model 3",
    "model_year": "2022"
  }
}

---

2. get_complaints

Retrieve consumer-submitted safety complaints for a vehicle.

Parameters:

Name	Type	Required	Description
make	string	Yes	Manufacturer name
model	string	No	Model name
model_year	string	No	Four-digit year
limit	integer	No	Max results (default 50, max 500)

Returns: List of complaint records with fields: odi_number, manufacturer, crash (bool), fire (bool), injuries, deaths, date_of_incident, date_filed, summary (truncated to 500 chars), components, make, model, model_year, source.

Example call:

{
  "name": "get_complaints",
  "arguments": {
    "make": "Ford",
    "model": "Bronco",
    "model_year": "2021",
    "limit": 20
  }
}

---

3. get_safety_ratings

Retrieve NHTSA 5-Star Safety Ratings for a vehicle.

Parameters:

Name	Type	Required	Description
make	string	Yes	Manufacturer name
model	string	No	Model name (returns all models if omitted)
model_year	string	Yes	Four-digit year

Returns: List of rating records with fields: vehicle_id, vehicle_description, overall_rating, front_crash_rating, side_crash_rating, rollover_rating, front_crash_driver_rating, front_crash_passenger_rating, make, model, model_year, source.

Example call:

{
  "name": "get_safety_ratings",
  "arguments": {
    "make": "Toyota",
    "model": "RAV4",
    "model_year": "2023"
  }
}

---

4. search_recalls_by_component

Search recalls across manufacturers by a component keyword. If no make is specified, searches across the 10 largest manufacturers (Toyota, Ford, Chevrolet, Honda, Nissan, BMW, Mercedes-Benz, Volkswagen, Hyundai, Kia).

Parameters:

Name	Type	Required	Description
component_keyword	string	Yes	Component to search (e.g. airbag, brake, steering)
make	string	No	Limit to one manufacturer. If empty, searches top 10 makes.
limit	integer	No	Max results (default 50, max 500)

Returns: List of matching recall records filtered by component.

Example call:

{
  "name": "search_recalls_by_component",
  "arguments": {
    "component_keyword": "Takata airbag",
    "limit": 100
  }
}

---

Running Modes

Batch Mode (default)

Set serveMcp: false (default). Provide make, model, modelYear, and limit as actor inputs. The actor fetches recalls for the specified vehicle and pushes results to the Apify dataset. Runs once and exits.

Input example:

{
  "make": "Toyota",
  "model": "Camry",
  "modelYear": "2020",
  "limit": 100
}

MCP Server Mode

Set serveMcp: true. The actor starts an HTTP server on port 4321 and stays alive for up to 24 hours. MCP clients can call any of the 4 tools in real time. The actor pushes a single seed record to the dataset for inspection.

Input example:

{
  "serveMcp": true,
  "make": "Ford"
}

---

MCP Endpoint Reference

Endpoint	Method	Description
/	GET	Health check — returns status and tool count
/health	GET	Health check
/mcp/tools	GET	List all available MCP tools with schemas
/mcp/call	POST	Call a tool by name with arguments

Request format for /mcp/call:

{
  "name": "get_recalls",
  "arguments": {
    "make": "Honda",
    "model": "Civic",
    "model_year": "2021"
  }
}

Response format:

{
  "content": [
    {
      "type": "text",
      "text": "[{\"recall_id\": \"...\", ...}]"
    }
  ]
}

---

Data Source

All data is sourced directly from the NHTSA DOT public API at https://api.nhtsa.dot.gov/. This is the authoritative US government database for:

• Recalls — Federal motor vehicle safety standard violations, manufacturer-initiated recalls, and NHTSA-ordered recalls going back decades.
• Complaints — Consumer-submitted safety complaints, including incidents involving crashes, fires, injuries, and deaths.
• Safety Ratings — Results from NHTSA's New Car Assessment Program (NCAP), the government's 5-star crash test program.

Data is fetched live on each actor run — no caching, always current.

---

Use Cases

Pre-Purchase Vehicle Research

Before buying a used or new vehicle, query recalls and safety ratings to understand the vehicle's safety history. Filter by year and model to see specific issues.

Fleet Safety Management

Organizations managing vehicle fleets can programmatically check all vehicles against current recall lists and prioritize remediation by units affected and severity.

Legal and Insurance Research

Attorneys and insurance analysts can query complaint data including crash/fire flags, injury counts, and death counts to identify patterns relevant to litigation or underwriting.

Automotive Journalism

Journalists covering automotive safety can monitor recall trends by component (e.g., Takata airbag scandal) across all major manufacturers.

AI Agent Integration

Connect this actor as an MCP server to Claude or another LLM. The agent can then answer questions like "Has the 2019 Honda CR-V been recalled for anything related to the fuel system?" by calling get_recalls directly.

Regulatory Compliance Monitoring

Legal and compliance teams can build automated checks that alert when new recalls are issued for company-owned or insured vehicles.

---

Error Handling

All tools follow the same graceful-failure pattern. On any error (network timeout, API unavailability, invalid parameters), the tool returns:

[{"_meta": {"error": "description of error", "fallback_tried": true}}]

The actor never crashes on API errors. It always exits with SUCCEEDED status, even when the upstream API is unavailable, ensuring integration test pipelines remain green.

---

Output Dataset Fields

Field	Type	Description
recall_id	string	NHTSA action number (unique recall identifier)
manufacturer	string	Vehicle manufacturer name
subject	string	Brief recall subject description
component	string	Affected vehicle component(s)
summary	string	Full recall description
consequence	string	Safety consequence if not remedied
remedy	string	Corrective action being taken
report_date	string	Date NHTSA received the recall report
units_affected	integer	Estimated number of vehicles affected
make	string	Vehicle make queried
model	string	Vehicle model queried
model_year	string	Model year queried
source	string	Always nhtsa.dot.gov

---

Technical Details

• Runtime: Python 3.11+
• Framework: Apify SDK v2+
• HTTP client: httpx (async)
• MCP server: asyncio raw TCP with HTTP/1.1 parsing (no external framework)
• Port: 4321
• Max server lifetime: 24 hours
• Timeout per API call: 20 seconds
• No API key required: NHTSA API is fully public

---

Connecting via Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "nhtsa-safety": {
      "url": "http://<your-actor-run-url>:4321"
    }
  }
}

Replace <your-actor-run-url> with the public URL of your running Apify actor instance.

---

License

Data sourced from NHTSA DOT is US government public domain data. This actor code is MIT licensed.