Promptly

Developer prompt analytics -- capture, analyze, and review AI conversations tied to development work.

Quick Start

npm i -g @getpromptly/cli
promptly init            # Configure AI tools + install /track skill
promptly start           # Prompts for ticket ID interactively
# ... work with any supported AI coding tool ...
promptly finish          # Save session data
promptly serve           # Open local dashboard at localhost:3000
promptly report          # View stats (select time period)
promptly digest          # Weekly insights digest with trends

Native /track Commands

After promptly init, use the /track command directly in your AI tool:

All tools support: /track <ticket-id>, /track status, /track finish

Features

• Local-first: Everything runs on your machine. No account needed.
• Multi-tool support: Works with Claude Code, Gemini CLI, Codex CLI, Cursor, Windsurf, and VS Code/Copilot via MCP.
• Native /track command: Built-in slash commands for Claude Code, Codex CLI, Gemini CLI, and VS Code + Copilot.
• Auto-prompt: AI tools offer session tracking at conversation start — no manual setup needed.
• Session tracking: Tag conversations to tickets, track token usage and duration.
• Git-aware sessions: Automatically captures commits, branches, and diff stats made during each session.
• Auto-categorization: Sessions are classified as bug-fix, feature, refactor, investigation, testing, or docs based on ticket ID, git commits, and conversation content.
• Session intelligence: Quality scoring (1-5 stars), tool usage analytics (Bash, Read, Edit, etc.), subagent tracking, context window metrics, and prompt quality analysis -- computed automatically at finish time.
• Session replay: Step through conversation turns with timing, playback controls, and cumulative stats to see how a session unfolded.
• Analytics dashboard: Cost-per-project trends, parallel session detection, skill usage analytics, instruction file effectiveness tracking, and aggregate prompt quality.
• Built-in dashboard: View sessions, analytics, digest, and session replay in your browser at localhost:3000.
• Optional cloud sync: For teams that want a shared dashboard.

Deployment Tiers

CLI Commands

How It Works

Claude Code / Gemini CLI / Codex CLI
        |
        └──> MCP Server --> ~/.promptly/buffer.json --> SQLite
                                                          |
                                            promptly serve (localhost:3000)

1. promptly init auto-detects installed AI coding tools and registers the MCP server with each.
2. promptly start TICKET-123 creates a session and signals the MCP server to begin logging.
3. As you work with your AI coding tool, the MCP server captures all conversation turns to a local buffer.
4. promptly finish captures git activity (commits, branch, diff stats, instruction file changes), auto-categorizes the session, computes session intelligence (quality score, tool usage, subagent stats, context window metrics, prompt quality analysis), writes data to SQLite, and clears the buffer.
5. promptly serve serves a dashboard with sessions, analytics, digest, and session replay.

Data Storage

All data is stored locally at ~/.promptly/:

Cloud Mode

For teams that want a shared dashboard, sign up at getpromptly.xyz.

promptly login             # Opens browser to sign in
promptly teams             # List your teams
promptly team set          # Interactive team selector (or: promptly team set my-team)
promptly start TICKET-1    # Creates session in your team
promptly finish            # Saves to SQLite AND uploads to API

Team Features

• Create teams at app.getpromptly.xyz/teams
• Invite members via shareable link (Owner/Admin can invite)
• Role-based access: Owner, Admin, Member
• Shared sessions: All team members can view sessions created within the team
• Team analytics: View usage by developer, model, and tool
• Per-tool comparison: See which AI coding tools your team uses most (Claude Code, Cursor, Gemini CLI, etc.)
• Data export: Download sessions and analytics as CSV or JSON from the dashboard

See docs/SETUP.md for cloud setup details.

Development

This is a TypeScript monorepo using Turborepo and pnpm.

pnpm install
pnpm build          # Build all packages
pnpm dev            # Dev mode (watch)

Packages

Documentation

• Setup Guide -- Installation and configuration
• Architecture -- Technical design and data flow

License

MIT