usa-npn-mcp-server
Enables querying and exporting phenology data from the USA National Phenology Network, including raw data, phenometrics, and mapping capabilities.
README
National Phenology Network MCP Server

Available MCP Tools
<!-- MCP-TOOLS-START -->
export-raw-data- Exports raw data to JSON or JSONL files in allowed directories.get-raw-data- Fetches raw data instead of summaries as from other tools. Limited to 300 records with truncation message if exceeded.individual-phenometrics- Fetches individual phenometrics (summarized data).magnitude-phenometrics- Fetches magnitude phenometrics (magnitude data).mapping- Maps site phenometrics onto a map of the USA with optional color labeling.observation-comment- Fetches observation comments based on observation_id.query-literature- Queries database of structured summaries from 175 papers that use phenology and phenometrics data.query-reference-material- Queries database containing NPN API reference material using a generated SQL query.site-phenometrics- Fetches site phenometrics (site-level data).status-intensity- Fetches status and intensity data (raw observation data). Use sparingly (can return massive datasets), prioritize phenometrics tools. <!-- MCP-TOOLS-END -->
Available MCP Resources
<!-- MCP-RESOURCES-START -->
recent-queries- List of recent query hash IDs and metadata for cached data access.available-roots- List of available root directories for file export operations. <!-- MCP-RESOURCES-END -->
Available MCP Prompts
<!-- MCP-PROMPTS-START -->
map-data- Generate a map of the data using the NPN API. <!-- MCP-PROMPTS-END -->
🧑🏿💻 Developing
Prerequisites
uv will manage the virtual environment and dependencies for you.
Install uv for your operating system:
<details> <summary>macOS and Linux</summary>
curl -LsSf https://astral.sh/uv/install.sh | sh
</details>
<details> <summary>Windows</summary>
In powershell, run:
# Install uv
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
</details>
Clone the repository
Using HTTPS (recommended for most users):
git clone https://github.com/VectorInstitute/usa-npn-mcp-server.git
Using SSH (if you have SSH keys configured with GitHub):
git clone git@github.com:VectorInstitute/usa-npn-mcp-server.git
After cloning with either method:
cd usa-npn-mcp-server
Installing dependencies
After uv is installed, run:
<details> <summary>macOS and Linux</summary>
uv sync
source .venv/bin/activate
</details>
<details> <summary>Windows</summary>
In powershell, run:
uv sync
. .\.venv\Scripts\activate.ps1
If the last command returns an error about running scripts being disabled on the system, you can run the following command in PowerShell to allow script execution for the current user:
Set-ExecutionPolicy RemoteSigned -Scope CurrentUser
Then run the activation command again:
. .\.venv\Scripts\activate.ps1
</details>
These commands set up and activate the .venv environment as specified in the pyproject.toml and uv.lock files.
Configuration
Configure for Claude Desktop App
You can download the Claude Desktop Application from here.
Once installed, you will need to modify your claude_desktop_config.json to make it aware of the MCP Server.
How to find and modify claude_desktop_config.json:
<details> <summary>macOS and Linux</summary>
- Open Claude Desktop app
- Click on "Claude" in the menu bar and select "Settings"
- In the Settings window, click on the "Developer" tab in the left sidebar
- Click the "Edit Config" button
- This will open a Finder window showing the location of the
claude_desktop_config.jsonfile - Open the file with your preferred text editor
<details> <summary>Add to claude config file:</summary>
{
"mcpServers": {
"npn": {
"command": "bash",
"args": [
"-c",
"source /absolute/path/to/usa-npn-mcp-server/.venv/bin/activate && uv run usa_npn_mcp_server /absolute/path/to/export/directory"
]
}
}
}
NOTE: Replace /absolute/path/to/usa-npn-mcp-server/ with local path to repo dir and /absolute/path/to/export/directory with local path to directory where you want exported files to be saved (or exclude this part to disable file export). You can specify multiple directories separated by spaces. For more info, see File Export Configuration.
</details> </details>
<details> <summary>Windows</summary>
- Open Claude Desktop app
- CTRL+Comma or Open the menu bar (three bar symbol top-left) and select "File" and "Settings"
- In the Settings window, click on the "Developer" tab in the left sidebar
- Click the "Edit Config" button
- This will open a window showing the location of the
claude_desktop_config.jsonfile - Open the file with your preferred text editor.
<details> <summary>Add to claude config file:</summary>
{
"mcpServers": {
"npn": {
"command": "cmd.exe",
"args": [
"/c",
"C:\\absolute\\path\\to\\usa-npn-mcp-server\\.venv\\Scripts\\activate.bat && uv run usa_npn_mcp_server C:\\absolute\\path\\to\\export\\directory"
]
}
}
}
NOTE: Replace C:\\absolute\\path\\to\\usa-npn-mcp-server\\ with local path to repo dir and C:\\absolute\\path\\to\\export\\directory with local path to directory where you want exported files to be saved (or exclude this part to disable file export). You can specify multiple directories separated by spaces - be sure to use backslashes for Windows paths. For paths containing spaces and more details, see File Export Configuration.
</details>
</details>
After saving the changes, restart Claude Desktop. If you receive no error messages from the Claude UI, the USA-NPN MCP server is likely installed correctly. Clicking the two buttons in the bottom left of the new chat box should reveal new options with npn labeling.

You should see a new local MCP server in the "Connectors" section of the Settings as below. This example has Filesystem and npn (usa-npn-mcp-server) MCP Servers enabled.

To enable tool use, you can click "Configure" (seen in the image above) and adjust which tools are enabled using the blue sliders and the permissions for each tool.

If you do not configure permissions this way, the tools will be enabled by default and permissions will be set to "Always ask permission" and each tool use will ask with an in-chat pop-up as below.

Recommended Complementary MCP Servers
Additional MCP Servers can be added to Claude Desktop in Settings using the Extensions tab by clicking Browse extensions. For use together with the USA-NPN MCP Server, the two following MCP Servers are recommended:
1. Filesystem:
- Let Claude access specified directories in your filesystem to read and write files.
- Configure
filesystemwith matching allowed directories asusa-npn-mcp-serverfor making data available during data analysis - both from recent queries saved to file and for externally sourced data files.
Note: A more fully featured alternative is Desktop Commander but it is not available in the listed extensions and would need to be configured by clicking add a custom one and performing custom installation and configuration. However, adding advanced or too many different MCP Servers to Claude can bloat your context window with tool descriptions. If you are experiencing context limitations, consider disabling some or all additional MCP Servers.
2. Context7:
- Let Claude access up-to-date code documentation for data analysis.
File Export Configuration
The server supports exporting data to files within specified directories. You can configure allowed export directories in two ways:
Command-line Arguments (Recommended)
Pass directory paths as arguments when starting the server:
# Single directory
uv run usa_npn_mcp_server /path/to/exports
# Multiple directories
uv run usa_npn_mcp_server /path/to/exports /path/to/another/dir
Environment Variable
Set the NPN_MCP_ALLOWED_DIRS environment variable:
# Unix/macOS (colon-separated)
export NPN_MCP_ALLOWED_DIRS="/path/to/exports:/path/to/another/dir"
# Windows (semicolon-separated)
set NPN_MCP_ALLOWED_DIRS="C:\path\to\exports;D:\another\dir"
Windows Paths with Spaces (Escaped Quotes Method)
For paths containing spaces, wrap each path with escaped double quotes (\"):
{
"mcpServers": {
"npn": {
"command": "cmd.exe",
"args": [
"/c",
"\\"C:\\\\Users\\\\John Doe\\\\Documents\\\\usa-npn-mcp-server\\\\.venv\\\\Scripts\\\\activate.bat\\" && uv run usa_npn_mcp_server \\"C:\\\\Program Files\\\\Data Export\\""
]
}
}
}
Multiple Windows Paths with Spaces
Quote each path individually when specifying multiple directories:
{
"mcpServers": {
"npn": {
"command": "cmd.exe",
"args": [
"/c",
"\\"C:\\\\Company Software\\\\NPN Tools\\\\usa-npn-mcp-server\\\\.venv\\\\Scripts\\\\activate.bat\\" && uv run usa_npn_mcp_server \\"D:\\\\Shared Data\\\\Research Projects\\" \\"E:\\\\Export Results\\\\Team Analysis\\""
]
}
}
}
Security Note: The server will only allow file operations within the specified directories for security.
Debugging
Debugging with MCP Inspector (Currently only macOS and Linux): To run a locally hosted MCP interpreter for debugging, use:
# Without file export
npx @modelcontextprotocol/inspector uv run usa_npn_mcp_server
# With file export to a specific directory
npx @modelcontextprotocol/inspector uv run usa_npn_mcp_server /path/to/exports
The first time you run this command you'll be prompted to download @modelcontextprotocol/inspector.
This command starts the MCP inspector within the uv-managed environment. The inspector can be used locally in-browser to inspect/test the server.
Testing dependencies: To install dependencies for testing (codestyle, unit tests, integration tests), run:
uv sync --dev
Other MCP Servers
For examples of other MCP servers and implementation patterns, see: https://github.com/modelcontextprotocol/servers
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。