opentrace

OpenTrace is a self-hosted observability server that exposes 75+ MCP tools for diagnosing and debugging production systems. It runs as a standalone Go binary and provides MCP-native workflows over stdio and SSE, allowing your AI assistant to query logs, traces, database stats, health checks, and more through guided tool suggestions. After starting the server, connect your MCP client (such as Claude Code, Cursor, or Windsurf) by pointing it at the opentrace process using a standard MCP config (the example in the README shows an opentrace entry with type stdio and the mcp command). The UI exposes dashboards for live logs and alerts, plus a taxonomy of tools organized by category (Log Intelligence, Database Introspection for Postgres, Analytics & Journeys, and more). Use the provided 75+ tools to diagnose issues, search logs, inspect traces, analyze database performance, and triage incidents. When your assistant loads, it will receive guided workflows and suggested_tools with pre-filled arguments to streamline the next steps in your debugging session.

Prerequisites:

• Go compiler installed (go1.20+ recommended)
• Git
• Optional: Docker if you prefer containerized deployment

From source (recommended for development):

1. Clone the repository: git clone https://github.com/adham90/opentrace.git cd opentrace

2. Build the binary: go build -o opentrace ./cmd/opentrace

3. Run the server: ./opentrace

4. Open the UI at http://localhost:8080 and follow the admin setup prompts.

Alternative (Docker):

• If you prefer Docker, pull and run the image (example): docker run -d --name opentrace -p 8080:8080 -v opentrace-data:/data ghcr.io/adham90/opentrace:latest

Optional configuration:

• Copy the environment example to configure environment variables used by OpenTrace: cp .env.example .env
• Inspect and set any necessary variables in .env (e.g., storage paths, ports, auth settings).

Usage example after install:

• Connect an MCP client by providing a config similar to the example in the README, pointing to the opentrace binary with the mcp command.

Tips and common notes:

• Default UI and API run on port 8080; adjust if needed via config or environment.
• OpenTrace is designed to be self-hosted with SQLite storage by default; ensure data directory is writable.
• When integrating with MCP clients (Claude Code, Cursor, Windsurf, etc.), use the stdio-based workflow defined in the config (opentrace mcp).
• The README mentions 75+ MCP tools; explore categories such as Log Intelligence, Database Introspection (Postgres), Errors, Analytics & Journeys, and more.
• If you need to customize tooling workflows, leverage the guided workflows where tools return suggested_tools with pre-filled arguments.
• For contribution or debugging, refer to the docs folder and the MCP integration guide linked in the README.