FBI Crime Data MCP Server
Provides access to FBI Crime Data Explorer API, enabling queries for crime statistics, arrests, hate crimes, NIBRS incidents, law enforcement employment, and more via 17 tools with smart caching and spillover handling.
README
FBI Crime Data MCP Server
An MCP (Model Context Protocol) server that provides access to the FBI's Crime Data Explorer API.
Query crime statistics, arrest data, hate crimes, NIBRS incidents, law enforcement employment, and more — directly from any MCP-compatible client.
Created for data journalists, researchers, and anyone interested in exploring US crime data with the power of language models. Initially created for use by the Policy Analyst Agent of the qsv Claude Cowork plugin.
Features
- 17 tools covering a wide range of crime data topics
- Crime trends and Summary Reporting System (SRS) crime data
- National Incident Based Reporting System (NIBRS) incident-based data and national estimates
- Arrest statistics with demographic breakdowns
- Hate crime incidents by bias motivation
- Expanded homicide and property crime details
- Police employment, Law Enforcement Officers Killed and Assaulted (LEOKA), Law Enforcement Suicide Data Collection (LESDC), and use of force
- Agency lookup, reference data, cache management, and spillover reading
- Geographic query levels — national, state, and agency for most tools; some also support region (
get_police_employment,get_nibrs_estimation) or agency-type / population-size breakdowns (get_nibrs_estimation) — all with automatic parameter validation - Smart yearly aggregation — monthly API data is automatically rolled up into yearly totals (sums for counts, averages for rates, last value for population), with an option for monthly granularity
- Tiered disk-backed caching — 90-day time-to-live (TTL) for stable data (trends, reference, summaries, NIBRS estimation), 30-day TTL for dynamic data (incidents, arrests, agency lookups), and 1-day TTL for the homepage summary (refresh dates change frequently)
- Spillover handling — responses exceeding 128K characters are saved to disk with a preview returned, so large queries are never silently truncated
- Input validation — date format/ordering checks, offense and bias code validation, and level-based parameter requirements with clear error messages
- Sliding-window rate limiting — 1,000 requests/hour with transparent wait-time feedback
- Reference tools for agency lookups (by state, Originating Agency Identifier (ORI), or district with name filtering) and code translations
Quick Start
-
Get a free API key from api.data.gov
-
Run with Claude Desktop — add to your
claude_desktop_config.json:
{
"mcpServers": {
"fbi-crime-data": {
"command": "uvx",
"args": ["fbi-crime-data-mcp"],
"env": {
"FBI_API_KEY": "your-api-key-here"
}
}
}
}
- Or run directly:
FBI_API_KEY=your-key uvx fbi-crime-data-mcp
Available Tools (17)
Core Crime Data
| Tool | Description |
|---|---|
get_summarized_crime_data |
SRS crime data — rates, actuals, clearances for violent crime, property crime, homicide, rape, robbery, assault, burglary, larceny, motor vehicle theft, arson |
get_nibrs_data |
NIBRS incident-based data for 70+ offense types |
get_arrest_data |
Arrest statistics by offense with optional demographic breakdowns (male, female, sex, race) |
get_crime_trends |
National crime trend percent changes across 10 crime types |
get_nibrs_estimation |
NIBRS national estimates by state, region, agency type, or population size |
Specialized Crime Data
| Tool | Description |
|---|---|
get_hate_crime_data |
Hate crime incidents by bias motivation (30+ categories) |
get_expanded_homicide_data |
Supplementary Homicide Reports — victim/offender demographics, weapons, circumstances |
get_expanded_property_data |
Expanded property crime details — stolen/recovered values for burglary, larceny, motor vehicle theft (MVT), robbery |
Law Enforcement Data
| Tool | Description |
|---|---|
get_police_employment |
Officer and civilian employee counts by gender, rates per 1,000 population |
get_leoka_data |
Officers killed and assaulted — weapons, circumstances, demographics |
get_lesdc_data |
Law enforcement suicide data — demographics, race, duty status, and more |
get_use_of_force_data |
Use of force incidents resulting in death, serious injury, or firearm discharge |
Overview
| Tool | Description |
|---|---|
get_cde_homepage_summary |
CDE homepage summary — mission statement, navigation, data freshness, date ranges, and national crime trends |
Reference & Lookup
| Tool | Description |
|---|---|
lookup_agency |
Find law enforcement agencies by state, ORI code, or judicial district |
get_reference_data |
State lists, offense/bias code lookups, data refresh dates |
manage_cache |
View cache stats, clear all entries, or clear only expired entries |
read_spillover |
Read spillover files saved when tool responses exceed the size limit |
Large Responses
When a tool response exceeds 128,000 characters, the full result is saved to ~/.cache/fbi-crime-data-mcp/spillover/ and a truncated preview is returned with the file path. To avoid this, narrow your query (shorter date range, specific state/agency).
Data Sources
All data comes from the FBI's Crime Data Explorer API, which provides Uniform Crime Reporting (UCR) data including both the Summary Reporting System (SRS) and the National Incident-Based Reporting System (NIBRS).
API Rate Limits
- Registered key: 1,000 requests per hour (rolling window)
- DEMO_KEY: 30 requests per IP per hour
The server includes a built-in rate limiter (1,000 req/hr). The DEMO_KEY limit is enforced API-side.
Development
# Install dependencies
uv sync
# Run the server locally
FBI_API_KEY=your-key uv run fbi-crime-data-mcp
# Run tests
uv run pytest
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。