JayantDevkar/claude-code-karma

Dashboard for monitoring claude code sessions.

What it solves

Claude Code stores session data (prompts, tool calls, and tokens) in local JSONL files that are difficult to read and are automatically deleted after 30 days. Claude Code Karma provides a local-first visual dashboard to browse, analyze, and organize this data without sending it to the cloud.

How it works

The project consists of a FastAPI backend that reads the ~/.claude/ directory and indexes metadata into a local SQLite database, and a SvelteKit frontend that visualizes the data. It can be installed as a desktop app for macOS and Windows to automate server management. For real-time monitoring, it uses Claude Code hooks to track live session states and can even trigger the OS window manager to bring the active terminal window to the front.

Who it’s for

Developers using the Claude Code CLI or Claude Desktop (in Claude Code mode) who want better visibility into their AI coding sessions, token costs, and tool usage.

Highlights

  • Session Analytics: Detailed breakdowns of token usage, costs, cache hit rates, and tool distribution per session.
  • Live Tracking: Real-time session status badges and a one-click button to focus the terminal window where a session is running.
  • Ticket Integration: Ability to link sessions to Linear, Jira, or GitHub Issues via MCP servers or branch name auto-detection.
  • Deep Inspection: Dedicated views for tasks, file operations, subagents, skills, and implementation plans.
  • Local-First: No cloud accounts or telemetry; all data stays on the user's machine.

Related

  • Project
    delexw/claude-code-trace

    A session log viewer and efficiency analyzer for Claude Code that transforms raw JSONL logs into readable, interactive conversations.

  • Project
    NoobyGains/claude-pulse

    A real-time usage monitor and status bar for Claude Code that tracks session limits, costs, and context usage without requiring an API key.

  • Project
    yakuikai/tokentab

    A local tool that aggregates session logs from AI coding assistants like Claude Code and Gemini CLI to track token usage and costs across projects.

  • Project
    nameforjt-afk/session-knowledge

    A local indexing tool for Claude Code that enables cross-project memory by making all session transcripts and credentials searchable via MCP tools.

  • Project
    soulduse/ai-token-monitor

    A system tray application for macOS and Windows that tracks real-time token usage and costs for AI coding tools by parsing local session logs.