Signoz

You can deploy and use the Signoz MCP Server to access Signoz data sources, dashboards, metrics, traces, and logs through a standardized MCP interface. This server lets you query and fetch Signoz content and run custom queries from your AI assistants or internal tools, enabling automation and intelligent observability workflows.

How to use

Connect your MCP client to the Signoz MCP Server using the HTTP/SSE endpoint or run the server locally and connect via a stdio-based workflow. The server exposes endpoints and functions that let you verify connectivity, list dashboards, fetch dashboard details and data, retrieve APM metrics, enumerate services, and run custom queries against ClickHouse or Signoz builder queries. When you issue commands from your MCP client, you’ll specify the server and the action you want (for example, test connectivity, fetch dashboards, or execute a builder query).

Practical usage patterns include: checking connectivity to ensure Signoz is reachable, listing dashboards to discover available views, fetching detailed metadata for a dashboard before rendering it, pulling panel data for a dashboard over a specific time range, and retrieving standard APM metrics for a service over a period. You can also execute custom SQL-like queries against ClickHouse or use Signoz builder queries to compute custom metrics and aggregations. If you need traces or logs, you can fetch those using the built-in data-type selector and limit parameters.

To integrate with an AI assistant or orchestration tool, point the MCP client at the local or remote MCP endpoint and issue the available actions in your workflow. You can run tests to verify connectivity, then progressively explore dashboards, services, and traces to assemble the data you need for dashboards, reports, or alerts.

How to install

Prerequisites: you need a runtime that can execute the MCP server and the ability to run either Python/uv or Docker Compose in your environment. You should have a running Signoz instance accessible from the MCP server and an optional API key if your Signoz instance requires one.

Option 1 – Install and run with uv (recommended for local development):

uv venv .venv
source .venv/bin/activate
uv sync

uv run -m src.signoz_mcp_server.mcp_server

Notes: ensure your config.yaml is in the same directory as mcp_server.py or supply the required environment variables as described in the configuration section.

Option 2 – Run with Docker Compose (recommended for production or containerized setups):

Edit the local configuration file at src/signoz_mcp_server/config.yaml with your Signoz details (host, API key if needed). Then start the server with Docker Compose.

# Start in detached mode
docker-compose up -d

Additional content

Configuration and health checks help you manage and validate the MCP server state. The MCP server reads connection settings from environment variables or a configuration file. You can override defaults at runtime to suit your deployment, including the Signoz host, API key, SSL verification, and server port.

Health check example: You can verify the server is responding on port 8000 with a simple health probe.

Available tools

test_connection

Verify connectivity to your Signoz instance and configuration.

fetch_dashboards

List all available dashboards from Signoz.

fetch_dashboard_details

Retrieve detailed information about a specific dashboard by its ID. Includes metadata about the dashboard.

fetch_dashboard_data

Fetch all panel data for a given dashboard by name and time range.

fetch_apm_metrics

Retrieve standard APM metrics (request rate, error rate, latency, apdex, etc.) for a given service and time range.

fetch_services

Fetch all instrumented services from Signoz with optional time range filtering.

execute_clickhouse_query

Execute custom Clickhouse SQL queries via the Signoz API with time range support.

execute_builder_query

Execute Signoz builder queries for custom metrics and aggregations with time range support.

fetch_traces_or_logs

Fetch traces or logs from Signoz using ClickHouse SQL. Specify data_type, time range, service, and limit.