Extract 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.
Local-first. No accounts. No cloud. No data leaves your machine.
Claude Code users: zero-config analysis, zero cost. Install the hook once. Every session gets analyzed automatically using your Claude subscription.
code-insights install-hook
# Try instantly (no install needed)
npx @code-insights/cli
# Or install globally
npm install -g @code-insights/cli
code-insights # sync sessions + open dashboard
code-insights install-hook # auto-sync + auto-analyze on session endThe dashboard opens at http://localhost:7890 and shows your sessions, analytics, and LLM-powered insights.
code-insights stats # terminal analytics (no dashboard needed)
code-insights stats today # today's sessions
code-insights dashboard # start dashboard server (auto-syncs first)
code-insights dashboard --no-sync # start dashboard without syncing
code-insights sync # sync sessions only
code-insights init # customize settings (optional)
code-insights doctor # diagnose your installation (start here if something's wrong)
| Tool | Data Location |
|---|---|
| Claude Code | ~/.claude/projects/**/*.jsonl |
| Cursor | Workspace storage SQLite (macOS, Linux, Windows) |
| Codex CLI | ~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl |
| Copilot CLI | ~/.copilot/session-state/{id}/events.jsonl |
| VS Code Copilot Chat | Platform-specific Copilot Chat storage |
Sessions from all tools are discovered automatically during sync.
code-insights dashboardOpens the built-in React dashboard at http://localhost:7890. The dashboard provides:
- Session Browser — global search (
Cmd+K), advanced filters (date range, outcome, saved presets), soft-delete, and full session details with chat view - Analytics — usage patterns, cost trends, activity charts
- LLM Insights — AI-generated summaries, decisions, learnings, and prompt quality analysis (7 deficit + 3 strength categories with dimension scores)
- 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)
- Export — LLM-powered cross-session synthesis in 4 formats (Agent Rules, Knowledge Brief, Obsidian, Notion)
- Settings — configure your LLM provider for analysis
code-insights dashboard --port 8080 # Custom port
code-insights dashboard --no-open # Start server without opening browser# Sync sessions and open dashboard — no setup required
code-insights
# Customize settings (optional) — prompts for Claude dir, excluded projects, etc.
code-insights init
# Show current configuration
code-insights config
# Configure LLM provider for session analysis (interactive)
code-insights config llm
# Configure LLM provider with flags (non-interactive)
code-insights config llm --provider anthropic --model claude-sonnet-4-20250514 --api-key sk-ant-...
# Show current LLM configuration
code-insights config llm --show
# Set a config value (e.g., disable telemetry)
code-insights config set telemetry false# Sync new and modified sessions (incremental)
code-insights sync
# Force re-sync all sessions
code-insights sync --force
# Preview what would be synced (no changes made)
code-insights sync --dry-run
# Sync only from a specific tool
code-insights sync --source cursor
code-insights sync --source claude-code
code-insights sync --source codex-cli
code-insights sync --source copilot-cli
# Sync only sessions from a specific project
code-insights sync --project "my-project"
# Quiet mode (useful for hooks)
code-insights sync -q
# Show diagnostic warnings from providers
code-insights sync --verbose
# Regenerate titles for all sessions
code-insights sync --regenerate-titles
# Soft-delete sessions (preview + confirm)
code-insights sync prune# Overview: sessions, cost, activity (last 7 days)
code-insights stats
# Cost breakdown by project and model
code-insights stats cost
# Per-project detail cards
code-insights stats projects
# Today's sessions with time, cost, and model details
code-insights stats today
# Model usage distribution and cost chart
code-insights stats models
# Cross-session patterns summary
code-insights stats patterns
Shared flags for all stats subcommands:
| Flag | Description |
|---|---|
--period 7d|30d|90d|all |
Time range (default: 7d) |
--project <name> |
Scope to a specific project (fuzzy matching) |
--source <tool> |
Filter by source tool |
--no-sync |
Skip auto-sync before displaying stats |
Cross-session pattern detection and synthesis. Requires an LLM provider to be configured.
# Generate weekly cross-session synthesis (current week)
code-insights reflect
# Reflect on a specific ISO week
code-insights reflect --week 2026-W11
# Scope to a specific project
code-insights reflect --project "my-project"
# Backfill facets for sessions that were synced before Reflect existed
code-insights reflect backfill
# Backfill prompt quality analysis
code-insights reflect backfill --prompt-qualityThe Reflect feature analyzes your sessions to surface:
- Friction points — recurring obstacles classified into 9 categories with attribution (user-actionable, AI capability, environmental)
- Effective patterns — working strategies across 8 categories with driver classification (user-driven, AI-driven, collaborative)
- Prompt quality — how well you communicate with AI tools (7 deficit + 3 strength categories)
- Working style — rules and skills derived from your sessions
# Check your installation — environment, database, providers, hooks, LLM, sync state
code-insights doctor
# Show what's wrong and fix it automatically (safe, idempotent fixes only)
code-insights doctor --fix
# Show probed paths for skipped/not-installed items
code-insights doctor --verbose
# Machine-readable output for bug reports
code-insights doctor --jsondoctor 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.
# Show sync statistics (sessions, projects, last sync)
code-insights status
# Open the local dashboard in your browser
code-insights open
code-insights open --project # Open filtered to the current project
# Delete all local data and reset sync state
code-insights reset --confirm# Analyze a session using your configured LLM provider
code-insights insights <session_id>
# Analyze using Claude Code (no API key needed — uses your Claude subscription)
code-insights insights <session_id> --native
# Check for unanalyzed sessions (last 7 days)
code-insights insights check
# Batch analyze all unanalyzed sessions
code-insights insights check --analyze
# Custom lookback window
code-insights insights check --days 14# Install Claude Code hooks — auto-sync + auto-analyze when sessions end
code-insights install-hook
# Install only the sync hook (no analysis)
code-insights install-hook --sync-only
# Install only the analysis hook
code-insights install-hook --analysis-only
# Remove all hooks
code-insights uninstall-hookAnonymous usage telemetry is opt-out. No PII is collected.
code-insights telemetry status # Check current status
code-insights telemetry disable # Disable telemetry
code-insights telemetry enable # Re-enable telemetryAlternatively, set the environment variable:
CODE_INSIGHTS_TELEMETRY_DISABLED=1 code-insights syncClaude Code users don't need to configure anything. Run code-insights install-hook and sessions are analyzed automatically using your Claude subscription.
For other tools, or if you prefer a different model, configure a provider via CLI or the dashboard Settings page:
code-insights config llmSupported providers:
| Provider | Models | Requires API Key |
|---|---|---|
| Claude Code (native) | Your Claude subscription model | No (via install-hook) |
| Anthropic | claude-opus-4-6, claude-sonnet-4-6, etc. | Yes |
| OpenAI | gpt-4o, gpt-4o-mini, etc. | Yes |
| Google Gemini | gemini-2.0-flash, gemini-2.0-pro, etc. | Yes |
| Ollama | llama3.2, qwen2.5-coder, etc. | No (local) |
API keys are stored in ~/.code-insights/config.json (mode 0o600, readable only by you).
This is a pnpm workspace monorepo with three packages: cli, dashboard, and server.
# Clone
git clone https://github.com/melagiri/code-insights.git
cd code-insights
# Install all dependencies
pnpm install
# Build all packages
pnpm build
# Link CLI for local testing
cd cli && npm link
code-insights --version
# Watch mode (CLI only)
cd cli && pnpm devcode-insights/
├── cli/ # This package — Node.js CLI, SQLite, providers
├── dashboard/ # Vite + React SPA
└── server/ # Hono API server (serves dashboard + REST API)
See CONTRIBUTING.md for code style, PR guidelines, and how to add a new source tool provider.
- All session data is stored in
~/.code-insights/data.db(SQLite) on your machine - No cloud accounts required
- No data is transmitted anywhere (unless you explicitly use an LLM provider with a remote API key)
- Anonymous telemetry collects only aggregate usage counts — no session content, no file paths
MIT — see LICENSE for details.