prometheus-mcp-server

# Prometheus MCP Server This is a simplified version of the Prometheus MCP server for querying metrics to provide AI analysis. ## Quick Start ### Launch 1. Set the Prometheus address environment variable: ```bash export PROMETHEUS_URL=your_Prometheus_endpoint ``` 2. Start the server: ```bash go run cmd/server/main.go sse ``` The server will start on port `:8081`, and can be accessed via the following URL: - http://localhost:8081/sse 3. (Optional) Set the log file path: ```bash export APP_LOG_FILE=/path/to/logfile.log ``` ### Testing ```bash npx @modelcontextprotocol/inspector node build/index.js ``` ## Features ### AI-Friendly Analysis Metrics 1. **JSON Format Output**: All query results are returned in JSON format, making it easy for AI systems to parse and process. ### MCP Tools 1. **search_metrics** - Search for metrics by regular expression - Parameters: `pattern` (metric name pattern, supports regular expressions) - Example: `http_.*` matches all metrics starting with `http_` - Returns: A list of matching metric names (in JSON format) 2. **get_metric_labels** - Get all labels for a metric - Parameters: `name` (metric name) - Example: `http_requests_total` - Returns: A list of all labels for the metric (in JSON format, excluding the `__name__` label) 3. **get_metric_label_values** - Get all values for a metric label - Parameters: - `name` (metric name) - `label` (label name) - Example: Get all values for the `method` label of the `http_requests_total` metric - Returns: A list of all values for the specified label (in JSON format) 4. **query** - Execute a PromQL instant query - Parameters: - `query` (PromQL query statement) - `time` (query timestamp, in RFC3339 format or Unix timestamp, optional) - Example: Query the current CPU usage - Returns: Query results (in JSON format) 5. **query_range** - Execute a PromQL range query - Parameters: - `query` (PromQL query statement) - `start` (start time, in RFC3339 format or Unix timestamp (milliseconds)) - `end` (end time, in RFC3339 format or Unix timestamp (milliseconds)) - `step` (query step, e.g., '15s', '1m') - Example: Query the CPU usage changes over the past hour, with one data point every 15 seconds - Returns: Range query results (in JSON format) 6. **query_chart** - Execute a PromQL range query and generate a chart - Parameters: - `query` (PromQL query statement) - `start` (start time, in RFC3339 format or Unix timestamp (milliseconds)) - `end` (end time, in RFC3339 format or Unix timestamp (milliseconds)) - `step` (query step, e.g., '15s', '1m') - `title` (chart title, optional) - Example: Query the CPU usage changes over the past hour and generate a chart - Returns: Chart image (in base64 encoded PNG format) ### Server Side 1. Set the Prometheus address: ```bash export PROMETHEUS_URL=your_Prometheus_endpoint ``` 2. Start the server: ```bash go run cmd/server/main.go sse ``` The server will expose Prometheus metrics on port `:8081` by default. ### MCP Client Configuration Configuration example: ```json { "mcpServers": { "products-sse": { "url": "http://localhost:8081/sse" } } } ``` ## Project Structure ``` . ├── cmd/ # Command line program entry │ └── server/ # Server code ├── internal/ # Internal implementation ├── pkg/ # Reusable packages │ └── prometheus # Prometheus related implementation └── go.mod # Go module definition ```