\n\nExtract decisions, learnings, and prompt quality scores from your AI coding sessions. Detect cross-session patterns. Get better at working with AI. Stores structured data in a local SQLite database and serves a built-in browser dashboard with LLM-powered synthesis.\n\n**Local-first. No accounts. No cloud. No data leaves your machine.**\n\n
\n \n
\n\n---\n\n> **Claude Code users: zero-config analysis, zero cost.**\n> Install the hook once. Every session gets analyzed automatically using your Claude subscription.\n> ```bash\n> code-insights install-hook\n> ```\n\n---\n\n## Quick Start\n\n```bash\n# Try instantly (no install needed)\nnpx @code-insights/cli\n\n# Or install globally\nnpm install -g @code-insights/cli\ncode-insights # sync sessions + open dashboard\ncode-insights install-hook # auto-sync + auto-analyze on session end\n```\n\nThe dashboard opens at `http://localhost:7890` and shows your sessions, analytics, and LLM-powered insights.\n\n### Individual commands\n\n```bash\ncode-insights stats # terminal analytics (no dashboard needed)\ncode-insights stats today # today's sessions\n\ncode-insights dashboard # start dashboard server (auto-syncs first)\ncode-insights dashboard --no-sync # start dashboard without syncing\ncode-insights sync # sync sessions only\ncode-insights init # customize settings (optional)\ncode-insights doctor # diagnose your installation (start here if something's wrong)\n```\n\n
\n \n
\n\n## Supported Tools\n\n| Tool | Data Location |\n|------|---------------|\n| **Claude Code** | `~/.claude/projects/**/*.jsonl` |\n| **Cursor** | Workspace storage SQLite (macOS, Linux, Windows) |\n| **Codex CLI** | `~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl` |\n| **Copilot CLI** | `~/.copilot/session-state/{id}/events.jsonl` |\n| **VS Code Copilot Chat** | Platform-specific Copilot Chat storage |\n\nSessions from all tools are discovered automatically during sync.\n\n## Dashboard\n\n```bash\ncode-insights dashboard\n```\n\nOpens the built-in React dashboard at `http://localhost:7890`. The dashboard provides:\n\n- **Session Browser** — global search (`Cmd+K`), advanced filters (date range, outcome, saved presets), soft-delete, and full session details with chat view\n- **Analytics** — usage patterns, cost trends, activity charts\n- **LLM Insights** — AI-generated summaries, decisions, learnings, and prompt quality analysis (7 deficit + 3 strength categories with dimension scores)\n- **Patterns** — weekly cross-session synthesis: friction points (with attribution), effective patterns (with driver classification), working style rules, and shareable AI Fluency Score card (downloadable 1200×630 PNG with score circle, fingerprint bars, and effective patterns)\n- **Export** — LLM-powered cross-session synthesis in 4 formats (Agent Rules, Knowledge Brief, Obsidian, Notion)\n- **Settings** — configure your LLM provider for analysis\n\n
\n \n
\n\n
\n \n
\n\n### Options\n\n```bash\ncode-insights dashboard --port 8080 # Custom port\ncode-insights dashboard --no-open # Start server without opening browser\n```\n\n## CLI Commands\n\n### Setup & Configuration\n\n```bash\n# Sync sessions and open dashboard — no setup required\ncode-insights\n\n# Customize settings (optional) — prompts for Claude dir, excluded projects, etc.\ncode-insights init\n\n# Show current configuration\ncode-insights config\n\n# Configure LLM provider for session analysis (interactive)\ncode-insights config llm\n\n# Configure LLM provider with flags (non-interactive)\ncode-insights config llm --provider anthropic --model claude-sonnet-4-20250514 --api-key sk-ant-...\n\n# Show current LLM configuration\ncode-insights config llm --show\n\n# Set a config value (e.g., disable telemetry)\ncode-insights config set telemetry false\n```\n\n### Sync\n\n```bash\n# Sync new and modified sessions (incremental)\ncode-insights sync\n\n# Force re-sync all sessions\ncode-insights sync --force\n\n# Preview what would be synced (no changes made)\ncode-insights sync --dry-run\n\n# Sync only from a specific tool\ncode-insights sync --source cursor\ncode-insights sync --source claude-code\ncode-insights sync --source codex-cli\ncode-insights sync --source copilot-cli\n\n# Sync only sessions from a specific project\ncode-insights sync --project \"my-project\"\n\n# Quiet mode (useful for hooks)\ncode-insights sync -q\n\n# Show diagnostic warnings from providers\ncode-insights sync --verbose\n\n# Regenerate titles for all sessions\ncode-insights sync --regenerate-titles\n\n# Soft-delete sessions (preview + confirm)\ncode-insights sync prune\n```\n\n### Terminal Analytics\n\n```bash\n# Overview: sessions, cost, activity (last 7 days)\ncode-insights stats\n\n# Cost breakdown by project and model\ncode-insights stats cost\n\n# Per-project detail cards\ncode-insights stats projects\n\n# Today's sessions with time, cost, and model details\ncode-insights stats today\n\n# Model usage distribution and cost chart\ncode-insights stats models\n\n# Cross-session patterns summary\ncode-insights stats patterns\n```\n\n
\n \n
\n\n**Shared flags for all `stats` subcommands:**\n\n| Flag | Description |\n|------|-------------|\n| `--period 7d\\|30d\\|90d\\|all` | Time range (default: `7d`) |\n| `--project ` | Scope to a specific project (fuzzy matching) |\n| `--source ` | Filter by source tool |\n| `--no-sync` | Skip auto-sync before displaying stats |\n\n### Reflect & Patterns\n\nCross-session pattern detection and synthesis. Requires an LLM provider to be configured.\n\n```bash\n# Generate weekly cross-session synthesis (current week)\ncode-insights reflect\n\n# Reflect on a specific ISO week\ncode-insights reflect --week 2026-W11\n\n# Scope to a specific project\ncode-insights reflect --project \"my-project\"\n\n# Backfill facets for sessions that were synced before Reflect existed\ncode-insights reflect backfill\n\n# Backfill prompt quality analysis\ncode-insights reflect backfill --prompt-quality\n```\n\nThe Reflect feature analyzes your sessions to surface:\n- **Friction points** — recurring obstacles classified into 9 categories with attribution (user-actionable, AI capability, environmental)\n- **Effective patterns** — working strategies across 8 categories with driver classification (user-driven, AI-driven, collaborative)\n- **Prompt quality** — how well you communicate with AI tools (7 deficit + 3 strength categories)\n- **Working style** — rules and skills derived from your sessions\n\n### Diagnostics\n\n```bash\n# Check your installation — environment, database, providers, hooks, LLM, sync state\ncode-insights doctor\n\n# Show what's wrong and fix it automatically (safe, idempotent fixes only)\ncode-insights doctor --fix\n\n# Show probed paths for skipped/not-installed items\ncode-insights doctor --verbose\n\n# Machine-readable output for bug reports\ncode-insights doctor --json\n```\n\n`doctor` checks ~30 things across 8 sections and tells you exactly what to run to fix each issue. If something isn't working, run this first. On a fresh install with no config, it shows a step-by-step setup guide instead of a check list.\n\n### Status & Maintenance\n\n```bash\n# Show sync statistics (sessions, projects, last sync)\ncode-insights status\n\n# Open the local dashboard in your browser\ncode-insights open\ncode-insights open --project # Open filtered to the current project\n\n# Delete all local data and reset sync state\ncode-insights reset --confirm\n```\n\n### Session Analysis\n\n```bash\n# Analyze a session using your configured LLM provider\ncode-insights insights \n\n# Analyze using Claude Code (no API key needed — uses your Claude subscription)\ncode-insights insights --native\n\n# Check for unanalyzed sessions (last 7 days)\ncode-insights insights check\n\n# Batch analyze all unanalyzed sessions\ncode-insights insights check --analyze\n\n# Custom lookback window\ncode-insights insights check --days 14\n```\n\n### Auto-Sync & Auto-Analyze Hook\n\n```bash\n# Install Claude Code hooks — auto-sync + auto-analyze when sessions end\ncode-insights install-hook\n\n# Install only the sync hook (no analysis)\ncode-insights install-hook --sync-only\n\n# Install only the analysis hook\ncode-insights install-hook --analysis-only\n\n# Remove all hooks\ncode-insights uninstall-hook\n```\n\n### Telemetry\n\nAnonymous usage telemetry is opt-out. No PII is collected.\n\n```bash\ncode-insights telemetry status # Check current status\ncode-insights telemetry disable # Disable telemetry\ncode-insights telemetry enable # Re-enable telemetry\n```\n\nAlternatively, set the environment variable:\n\n```bash\nCODE_INSIGHTS_TELEMETRY_DISABLED=1 code-insights sync\n```\n\n## LLM Configuration\n\n**Claude Code users don't need to configure anything.** Run `code-insights install-hook` and sessions are analyzed automatically using your Claude subscription.\n\nFor other tools, or if you prefer a different model, configure a provider via CLI or the dashboard Settings page:\n\n```bash\ncode-insights config llm\n```\n\n**Supported providers:**\n\n| Provider | Models | Requires API Key |\n|----------|--------|-----------------|\n| Claude Code (native) | Your Claude subscription model | No (via `install-hook`) |\n| Anthropic | claude-opus-4-6, claude-sonnet-4-6, etc. | Yes |\n| OpenAI | gpt-4o, gpt-4o-mini, etc. | Yes |\n| Google Gemini | gemini-2.0-flash, gemini-2.0-pro, etc. | Yes |\n| Ollama | llama3.2, qwen2.5-coder, etc. | No (local) |\n\nAPI keys are stored in `~/.code-insights/config.json` (mode 0o600, readable only by you).\n\n## Development\n\nThis is a pnpm workspace monorepo with three packages: `cli`, `dashboard`, and `server`.\n\n```bash\n# Clone\ngit clone https://github.com/melagiri/code-insights.git\ncd code-insights\n\n# Install all dependencies\npnpm install\n\n# Build all packages\npnpm build\n\n# Link CLI for local testing\ncd cli && npm link\ncode-insights --version\n\n# Watch mode (CLI only)\ncd cli && pnpm dev\n```\n\n### Workspace Structure\n\n```\ncode-insights/\n├── cli/ # This package — Node.js CLI, SQLite, providers\n├── dashboard/ # Vite + React SPA\n└── server/ # Hono API server (serves dashboard + REST API)\n```\n\n### Contributing\n\nSee [CONTRIBUTING.md](https://github.com/melagiri/code-insights/blob/master/CONTRIBUTING.md) for code style, PR guidelines, and how to add a new source tool provider.\n\n## Privacy\n\n- All session data is stored in `~/.code-insights/data.db` (SQLite) on your machine\n- No cloud accounts required\n- No data is transmitted anywhere (unless you explicitly use an LLM provider with a remote API key)\n- Anonymous telemetry collects only aggregate usage counts — no session content, no file paths\n\n## License\n\nMIT — see [LICENSE](https://github.com/melagiri/code-insights/blob/master/LICENSE) for details.\n","readmeFilename":"README.md"}
![\"Code](\"https://raw.githubusercontent.com/melagiri/code-insights/master/docs/assets/logo.svg\"){kind=logo}
\n![\"AI](\"https://raw.githubusercontent.com/melagiri/code-insights/master/docs/assets/screenshots/code-insights-ai-fluency-score.png\")
\n![\"Session](\"https://raw.githubusercontent.com/melagiri/code-insights/master/docs/assets/screenshots/session-insight-light.png\")
\n![\"Patterns](\"https://raw.githubusercontent.com/melagiri/code-insights/master/docs/assets/screenshots/patterns-light.png\")
\n![\"Analytics](\"https://raw.githubusercontent.com/melagiri/code-insights/master/docs/assets/screenshots/analytics-light.png\")
\n![\"Terminal](\"https://raw.githubusercontent.com/melagiri/code-insights/master/docs/assets/screenshots/stats.png\")
\n