API observability tools can't see your subscription spend.
Helicone, Langfuse, Portkey — they track tokens. That $20/seat Cursor and $19/seat Copilot your team runs? Invisible. Stax tracks both.
live in different ledgers
Your AI bill is everywhere — OpenAI and Anthropic APIs, Cursor and Copilot seats, ChatGPT Team. Most cost trackers see the API portion and miss the subscription half entirely. Stax pulls all of it into one number. $49/month flat. No code changes. Built for the person who pays the bill, not the engineer who writes the prompts.
Helicone, Langfuse, Portkey — they track tokens. That $20/seat Cursor and $19/seat Copilot your team runs? Invisible. Stax tracks both.
Founders, CFOs, and CTOs need cost clarity — not a dashboard full of "tokens," "traces," and SDK config. Stax shows dollars, by team, by month.
Usage spikes, observability bill spikes. Every existing tool charges more exactly when you need it most. Stax is $49/mo regardless of how much you spend on AI.
OpenAI, Anthropic, Cursor, and Copilot sync nightly via admin API — paste a key, that's it. For vendors without admin APIs yet (Google AI, Mistral, ChatGPT Business, Perplexity), drop your monthly invoice and we parse it in 30 seconds. When those vendors ship APIs, we flip them to live.
Tag spend by team, project, person, or environment. Every dollar is attributed automatically — no instrumentation, no code changes, no engineer required.
Per-provider or rollup budgets. Alerts at 80% and 100% land in Slack or email — before the surprise invoice does.
Connect up to 3 providers. No card, no time limit.
Unlimited providers. 12 months of history. Board-ready exports.
Everything in Pro, plus per-seat attribution and SSO.
Those tools are observability platforms — they trace API calls and help you debug prompts. Stax is a finance tool. It tracks every dollar across APIs and subscriptions: the $20/seat/mo Cursor your team runs, the $25/seat/mo ChatGPT Team, the Copilot charges — none of that appears in any observability dashboard. If you want to know what your AI stack costs, use Stax. If you want to debug why a prompt misfired, Langfuse or Portkey are built for that (note: Helicone entered maintenance mode in March 2026 after being acquired by Mintlify).
Correct. For those vendors we use a monthly invoice upload — drop the PDF and we parse it in about 30 seconds. It's a manual step once a month until those vendors ship admin APIs. When they do, we flip them to live sync automatically.
It doesn't change. $49/mo covers unlimited providers and unlimited spend volume. You're not penalized for using more AI. The Team plan at $99/mo adds per-seat attribution, SSO, and audit logs — but even that is flat, not per-seat.
We read billing metadata — never prompt content or model responses. Keys are encrypted at rest. Full detail is on the security page. SOC 2 Type II is in progress.
No. We connect to billing endpoints and invoice metadata only. Your prompts, model responses, and request payloads never touch our servers.
Cost-control software shouldn't be a variable cost. If your AI bill triples next month, your Stax bill doesn't move.
Yes — 3 providers, 30 days of history, no card required. Most teams upgrade when they hit provider #4 or need more than a month of history.
Email us at [email protected]. We add the top-requested vendors regularly. In the meantime, drop your invoice and we'll track it as a flat-cost line item.
Start free — 3 providers, 30 days of history, no card required. Upgrade when you need more.