fred-mcp
Enables querying and retrieving Federal Reserve Economic Data (FRED) including series, categories, releases, sources, and tags, with support for stdio and HTTP transports and bring-your-own-key authentication.
README
Federal Reserve Economic Data MCP Server
[!NOTE] This open-source project is not affiliated with, sponsored by, or endorsed by the Federal Reserve or the Federal Reserve Bank of St. Louis. "FRED" is a registered trademark of the Federal Reserve Bank of St. Louis, used here for descriptive purposes only.
A production-grade Model Context Protocol server for FRED® economic data. Covers all 31 endpoints exposed by fred-py-api.
Features
- Full FRED API v1 coverage: series, categories, releases, sources, and tags
- FastMCP 3.x with structured tool output and proper
ToolErrorhandling - Dual-mode credentials: environment variable (stdio) or per-client HTTP header (remote BYOK)
- Transports:
stdio,streamable-http, andsse - Docker image published to GHCR
Installation
pip install fred-mcp
Requires Python 3.10+.
Get a free FRED API key at fredaccount.stlouisfed.org/apikey.
Setup
Remote hosted server (recommended)
Use the public hosted server with bring your own key (BYOK): no install and no shared server-side API key. Each client sends its own FRED API key in the X-FRED-API-Key header.
Transport: streamable HTTP. Endpoint: https://fred-mcp-prod.fly.dev/mcp.
Add this to your MCP client's configuration file and restart the client:
{
"mcpServers": {
"fred-mcp": {
"url": "https://fred-mcp-prod.fly.dev/mcp",
"headers": {
"X-FRED-API-Key": "<your fred api key>"
}
}
}
}
Programmatic example with the FastMCP client:
from fastmcp import Client
from fastmcp.client.transports import StreamableHttpTransport
transport = StreamableHttpTransport(
"https://fred-mcp-prod.fly.dev/mcp",
headers={"X-FRED-API-Key": "your_fred_api_key"},
)
async with Client(transport=transport) as client:
await client.ping()
Local (stdio)
For clients that spawn a local process, install fred-mcp and use:
{
"mcpServers": {
"fred-mcp": {
"command": "fred-mcp",
"env": {
"FRED_API_KEY": "<your fred api key>"
}
}
}
}
Or run directly from the terminal:
export FRED_API_KEY=your_api_key
fred-mcp
Self-hosted HTTP
Deploy your own instance and use the same url + headers configuration, substituting your host for the endpoint above. No shared server-side key is required when clients send X-FRED-API-Key.
For public internet deployment, terminate TLS at a reverse proxy (nginx, Caddy, Cloudflare). Do not expose plain HTTP with API keys.
Docker
Run the server:
docker run -d -p 8000:8000 \
--name fred-mcp-server \
ghcr.io/zachspar/fred-mcp/fred-mcp-server:latest
Connect with your FRED API key in the X-FRED-API-Key header (same JSON shape as the remote hosted example, with url set to your deployment).
For stdio via Docker:
{
"mcpServers": {
"fred-mcp": {
"command": "docker",
"args": [
"run", "-i", "--rm",
"-e", "MCP_SERVER_TRANSPORT=stdio",
"-e", "FRED_API_KEY=<your fred api key>",
"ghcr.io/zachspar/fred-mcp/fred-mcp-server:latest"
]
}
}
}
Optional server-side fallback key
Set FRED_API_KEY on the server for clients that cannot send custom headers. Header takes precedence when both are present.
| Variable | Default | Description |
|---|---|---|
FRED_API_KEY |
— | FRED API key (required for stdio; optional HTTP fallback) |
FRED_API_KEY_HEADER |
X-FRED-API-Key |
HTTP header name for BYOK |
MCP_SERVER_TRANSPORT |
stdio |
stdio, streamable-http, or sse |
MCP_SERVER_HOST |
localhost |
Bind host for HTTP transports |
MCP_SERVER_PORT |
8000 |
Bind port for HTTP transports |
Tools (31)
Series
| Tool | Description |
|---|---|
get_series |
Series metadata |
get_series_categories |
Categories for a series |
get_series_observations |
Data values / observations |
get_series_release |
Release that publishes a series |
search_series |
Search series by text or ID |
search_series_tags |
Tags for a series search |
search_series_related_tags |
Related tags for a series search |
get_series_tags |
Tags on a series |
get_series_updates |
Recently updated series |
get_series_vintage_dates |
Vintage / revision dates |
Categories
| Tool | Description |
|---|---|
get_category |
Category metadata |
get_category_children |
Child categories |
get_category_related |
Related categories |
get_category_series |
Series in a category |
get_category_tags |
Tags in a category |
get_category_related_tags |
Related tags in a category |
Releases
| Tool | Description |
|---|---|
list_releases |
All releases |
list_release_dates |
Release dates across releases |
get_release |
Release metadata |
get_release_dates |
Dates for one release |
get_release_series |
Series in a release |
get_release_sources |
Sources for a release |
get_release_tags |
Tags for a release |
get_release_related_tags |
Related tags for a release |
get_release_tables |
Release tables |
Sources
| Tool | Description |
|---|---|
list_sources |
All data sources |
get_source |
Source metadata |
get_source_releases |
Releases from a source |
Tags
| Tool | Description |
|---|---|
list_tags |
Search / list tags |
get_related_tags |
Related tags |
get_tags_series |
Series matching tags |
Migration from 0.x
Version 1.0.0 renames tools to snake_case and upgrades to FastMCP 3.x.
| Old name (0.x) | New name (1.0) |
|---|---|
FREDSeries |
get_series |
FREDSeriesCategories |
get_series_categories |
FREDSeriesObservations |
get_series_observations |
FREDSeriesRelease |
get_series_release |
FREDSeriesSearch |
search_series |
FREDSeriesSearchTags |
search_series_tags |
FREDSeriesSearchRelatedTags |
search_series_related_tags |
FREDSeriesTags |
get_series_tags |
FREDSeriesUpdates |
get_series_updates |
FREDSeriesVintageDates |
get_series_vintage_dates |
21 additional tools were added for categories, releases, sources, and tags.
Error responses now use MCP ToolError (isError: true) instead of {"error": ...} payloads.
Development
python3.13 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
pytest
ruff check src tests
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。
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。