AI Observer is a self-hosted observability backend tailored for monitoring AI coding tools like Claude Code, Gemini CLI, and OpenAI Codex CLI. It's designed to provide real-time insights into token usage, costs, API latency, error rates, and session activity through a unified dashboard. The tool runs as a single binary, eliminating the need for external dependencies, which simplifies setup and maintenance. The dashboard features customizable widgets and real-time updates via WebSocket. Users can import historical data from local JSONL or JSON files, enabling cost tracking for over 67 models across the supported tools. The storage utilizes DuckDB for fast analytics on large datasets. Installation is straightforward, whether through Docker or Homebrew, and it supports multiple architectures. AI Observer includes commands for data import, export, and deletion, allowing users to manage telemetry efficiently. For instance, users can export telemetry data to Parquet files and even create ZIP archives of their exports. It also supports detailed filtering options for both import and export, letting users specify date ranges and service types. Environment variables facilitate configuration, including enabling telemetry and setting up exporters.