mcp-swarmia
Exposes the Swarmia Export API as tools for AI assistants to query engineering metrics through natural language. It enables retrieval of DORA metrics, pull request cycle times, investment balances, and FTE effort reports.
README
mcp-swarmia
A local Model Context Protocol (MCP) server that exposes the Swarmia Export API as tools for AI assistants such as Claude.
Once connected, Claude (or any MCP-compatible client) can query your engineering metrics—pull request cycle times, DORA metrics, investment balance, CapEx reports, and FTE effort—in natural language without you having to write a single API call.
Requirements
- Python 3.11+
- A Swarmia account with an API token (Settings → API tokens)
mcpPython package (only external dependency)
Installation
# 1. Clone or copy this directory
cd mcp-swarmia
# 2. Create and activate a virtual environment (recommended)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
# 3. Install dependencies
pip install -r requirements.txt
Configuration
Export your Swarmia API token as an environment variable:
export SWARMIA_API_TOKEN="your-token-here"
Alternatively, add it to a .env file and source it, or configure it in your MCP client settings (see below).
Running the server
python server.py
The server communicates over stdio (standard input/output), which is the transport expected by Claude Desktop and most other MCP clients. You do not need to expose any network ports.
Connecting to Claude Desktop
Add the following block to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"swarmia": {
"command": "/absolute/path/to/.venv/bin/python",
"args": ["/absolute/path/to/mcp-swarmia/server.py"],
"env": {
"SWARMIA_API_TOKEN": "your-token-here"
}
}
}
}
Replace the paths and token with your actual values, then restart Claude Desktop.
Available tools
| Tool | Description | Required params |
|---|---|---|
get_pull_requests |
Team-level PR metrics (cycle time, review time, size, …) | — |
get_dora_metrics |
Organisation DORA metrics (deployment frequency, lead time, CFR, MTTR) | — |
get_investment_balance |
Monthly FTE investment balance by category | startDate, endDate |
get_capex |
Software capitalisation report | startDate, endDate |
get_capex_employees |
Per-employee CapEx breakdown | year |
get_fte |
Monthly engineering effort (FTE) by author | month |
Shared optional parameters
All time-series endpoints accept:
| Parameter | Description | Example |
|---|---|---|
timeframe |
Preset window | last_30_days |
startDate |
Custom start (YYYY-MM-DD) | 2024-01-01 |
endDate |
Custom end (YYYY-MM-DD) | 2024-03-31 |
timezone |
tz database identifier | America/New_York |
timeframe and startDate/endDate are mutually exclusive. Defaults to last_7_days when neither is provided.
DORA-specific parameters
| Parameter | Description |
|---|---|
app |
Filter by deployment application name |
environment |
Filter by deployment environment |
FTE-specific parameters
| Parameter | Description | Values |
|---|---|---|
groupBy |
How to group effort | highestLevelIssue (default), lowestLevelIssue, customField |
customField |
Jira field ID | Required when groupBy=customField |
Response format
All tools return CSV text with a header row and comma delimiters, exactly as returned by the Swarmia API. No summary aggregations are added.
Example prompts
After connecting the server to Claude, try:
- "Show me our DORA metrics for the last 30 days."
- "What does our pull request cycle time look like for Q1 2024?"
- "Fetch the investment balance for January 2024."
- "Get the CapEx report for 2024."
- "Show FTE effort for March 2024, grouped by highest-level issue."
Security note
Swarmia recommends passing tokens via the Authorization header rather than query parameters, as some proxies log query strings. This server always uses the header approach.
Troubleshooting
| Problem | Solution |
|---|---|
SWARMIA_API_TOKEN is not set |
Export the env var or add it to your MCP client config |
HTTP 401 |
Token is invalid or expired — regenerate it in Swarmia |
HTTP 400 |
Check that required parameters (e.g. startDate/endDate) are provided and in the correct format |
| Server not appearing in Claude | Verify the absolute paths in claude_desktop_config.json and restart Claude Desktop |
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。
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。
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。