Observe Community

You gain an MCP server that lets LLMs access Observe platform data through semantic search, automatic dataset discovery, and metrics intelligence. This setup enables natural language queries to discover data sources, understand metrics, and run OPAL queries across Observe datasets with contextual guidance and live documentation search.

How to use

Use the MCP server with your preferred MCP client to ask natural language questions about Observe datasets, metrics, and documentation. You can discover datasets by description, learn OPAL syntax, run queries against one or more datasets, and receive intelligent guidance when errors occur. The system provides live documentation grounding, auto-discovery of datasets and metrics, and structured results designed for seamless integration with your language model workflow.

How to install

Prerequisites include Docker with Docker Compose and Python 3.13 or newer for manual setup.

1. Clone the MCP server repository and navigate into the project directory.

2. Copy and configure environment variables.

3. Start the services with Docker Compose.

Other important content

Configuration and security details are provided below for immediate use. The server supports HTTP and local stdio configurations, including authentication, intelligence data loading, and maintenance tasks. It includes OpenTelemetry collection for telemetry data and a Gemini AI-powered live documentation search grounding.

Environment variables shown in examples include credentials for the Observe platform, public keys for MCP authentication, and tokens for OpenTelemetry collection. Use placeholder values where you replace them with your own credentials.

OpenTelemetry integration is available via a collector that forwards telemetry to Observe with proper authentication, retry semantics, and resource attributes. The collector exposes OTLP endpoints for gRPC and HTTP and includes a health check endpoint.

For production deployments, you can place the MCP server behind an HTTPS proxy with SSL/TLS and refer to a secure remote URL for client configuration.

Authentication

JWT-based authentication protects access to the MCP server. You generate RSA keys, publish the public key to initialize authentication, and create user tokens with defined permissions. On successful login, the server assumes the identity and permissions of the configured Observe API token.

Maintenance

Regular maintenance includes refreshing intelligence data for datasets and metrics, monitoring server logs and database status, and validating token configurations. You should run intelligence updates when new datasets or metrics appear and check the Gemini AI documentation search usage for ongoing indexing health.

Examples and integration notes

To connect a client, you can configure the MCP server URL and authorization header in your client settings. The server supports multiple connection methods, including remote HTTP URLs and local stdio execution commands.

{
  "mcpServers": {
    "observe": {
      "type": "http",
      "url": "http://localhost:8000/mcp",
      "headers": {
        "Authorization": "Bearer your_mcp_token_here"
      }
    }
  }
}

Available tools

discover_context

Unified discovery tool for datasets and metrics showing dimensions, schemas, and query templates in one search to reduce dimension guessing

learn_observe_skill

Search OPAL skill documentation using local BM25 indexing for OPAL syntax and best practices without external API dependencies

execute_opal_query

Run OPAL queries across single or multiple Observe datasets with robust error handling and structured results