See what your AI agents
are actually doing.

Every developer on your team talks to AI all day. Right now, you have no idea what they're asking it, whether the prompts are any good, or what it costs. Promptly logs every session and puts it on a dashboard. You see who's burning through Opus on formatting tasks, who wrote the prompt worth copying, and where the budget actually went.

$ npm i -g @getpromptly/cli
Start Free Trial Try the Team Dashboard View on GitHub

Works with Claude Code, Codex CLI, Gemini CLI, VS Code + Copilot, Cursor, and Windsurf. Everything stays on your machine unless you turn on the team dashboard.

Terminal
| PR MPTLY |
My Sessions Acme Team
0
Sessions
$0
Est. Cost
0m
Total Time
Monthly Budget $892 / $1,250
Team members
AK
SJ
MR
+2
FEAT-142 AKAlex opus-4 $3.42 ★
BUG-89 SJSarah gpt-4o $1.87 ☆
FEAT-143 MRMike sonnet $0.94 ☆

What it does

Know what your team is doing

  • Sarah spent 3 hours on BUG-89 with GPT-4o. Mike finished FEAT-143 in 20 minutes with Sonnet. You can see both.
  • Every session is logged with the developer, the ticket, the model, and the cost
  • Star the sessions that worked well and share them across the team

Prompt review is the new code review

  • Teams review code before it ships. They should review the prompts that produced it, too.
  • promptly review scores a prompt on clarity, context, and cost before the agent runs
  • Catches vague prompts that waste 3 turns of back-and-forth, and suggests a rewrite

See where the money goes

  • Dollar cost per session, per developer, per ticket. No more guessing.
  • Set monthly budgets. Get an alert before someone blows past them.
  • Spots the obvious waste: Opus for a one-line fix, the same boilerplate prompt run 40 times

Review before running

promptly review tells you if the prompt is good enough before it burns tokens

Find the waste

promptly optimize spots repeated patterns and prompts that should be one-liners

Works with everything

Claude Code, Codex CLI, Gemini CLI, VS Code, Cursor, Windsurf

Stays on your machine

Prompts, code, and file paths are stored locally. Cloud is opt-in.

How it works

promptly init # auto-detects your AI tools
promptly start TICKET-1 # starts capturing the session
# ... code with any AI tool ...
promptly finish # calculates cost, saves to dashboard

✓ Session saved: 24,521 tokens · $3.42 · 47m

Security & Privacy

Data isolation

Your sessions are yours. Team sessions are only visible to team members with strict access controls.

Role-based access

Owner, Admin, Member roles. Control who can invite, manage, or view team data.

Your code stays private

Conversation data is isolated per-user and per-team. No cross-tenant data access.

Local-first option

Keep everything on your machine. SQLite database, no cloud required. Full offline support.

Pricing

Free

$0
  • promptly review (bring your own Anthropic key)
  • promptly optimize (no API key needed)
  • Unlimited sessions
  • Cost estimates (250+ models)
  • Local dashboard
  • Local-first — nothing leaves your machine

Team

$30$15/month
50% off — early access
  • All Free features, plus:
  • Cloud dashboard with per-developer breakdowns
  • Monthly token budgets + threshold alerts
  • Session bookmarking + team starred feed
  • Tickets tied to JIRA / Linear issues
  • Unlimited team members
  • 14-day free trial
Start Free Trial