Agencies burn budget across a dozen AI tools and APIs, with no idea which client it's for. SpendGuard is the control tower: track every call and subscription, attribute it to the right client, and rebill with margin.
14-day free trial · No credit card to start · Set up in minutes
One line shows what your Claude actually costs (models, MCP tools, agents) and signs you in from the terminal. Nothing to configure.
npx spendguardco claude
One big number per provider, per month. No breakdown by person, by client, or by project. By the time you see it, the money is gone.
Which client did this €4.200 OpenAI bill actually serve? Nobody can say.
How many AI subscriptions are we paying for, and who's still using them?
We could rebill AI cost to clients with margin, if only we could measure it.
Three things, in one place, built for how agencies actually work.
A proxy sits between your team and the AI APIs. Every call is logged with who, what, when, and exactly how much, across OpenAI, Anthropic, Runway, ElevenLabs and more. Budgets stop runaways before they hit the invoice.
Per-user · Per-provider · Real-timeTrack every fixed AI subscription, ChatGPT Team, Midjourney, Runway, Canva, with seats, renewal dates and monthly cost. SpendGuard combines subscriptions and API spend into one total, and warns you before renewals.
Seats · Renewals · Total costTag every call to a client, project or deliverable, by key scope or a request header. Then run rebilling reports: cost, your markup, and the billable amount per client, exportable to CSV. This is the part other tools don't do.
Attribution · Markup · CSV exportSpendGuard is the only AI spend tool built around per-client rebilling. Set a markup, pick a period, and get a clean breakdown ready for the client invoice.
Five places where agency money leaks, and what SpendGuard does about each. Numbers are typical for a 10 to 20 person agency.
An agent that keeps running costs that before anyone notices. SpendGuard stops it at the budget you set and emails what happened within minutes. One caught loop pays for years of Agency.
Half the team on the priciest model for tasks a small one does at a tenth of the cost. On €2.000 API spend, moving 30% to cheaper models is €600 a month, every month.
ChatGPT Team, Midjourney, Runway, ElevenLabs: €20 to €60 seats for people who left or stopped using them. Two forgotten seats is already €100 a month.
Every euro of AI spent for a client sits on your bill. SpendGuard turns it into an invoice line. €4.000 of client AI cost rebilled at 25% is €1.000 of margin a month that does not exist right now.
Plus two hours a month finance no longer spends reconstructing eight AI invoices per client. Between €500 and €1.500 a month back, for €29 or €79.
Free covers what every gateway gives away. You pay for the part nobody else does: knowing what AI costs per client, and billing it back.
Start with 14 days of everything, then keep Free or pick a plan. Prices in EUR excl. VAT, billed monthly. No credit card to start.
"We built SpendGuard because we were running AI across two agencies and flying blind. It's the dashboard I wished existed, so we made it." — Meinte Stinstra, Founder, JackandAI
Your team's tools point at SpendGuard's proxy instead of the provider directly, via a base URL and a SpendGuard key, the npx spendguardco command, or the spendguardco Python package. SpendGuard forwards the call to the real provider, logs the cost, and returns the response. Your real provider keys stay encrypted on our side and are never exposed to your team.
Usually just two environment variables, or one line with the Python SDK (pip install spendguardco, then spendguardco.guard()). For OpenAI, Anthropic and Gemini it is a base URL swap; npx spendguardco configure rewrites your .env for you. Streaming, tool use and the rest keep working.
Two ways. Scope a proxy key to a client, project or deliverable so everything through that key is attributed automatically. Or pass an X-SpendGuard-Client header per request when one key serves many clients. The rebilling report then rolls cost up per client with your markup.
OpenAI and Anthropic with full streaming, plus Runway, ElevenLabs, Replicate, Stability, Fal, HeyGen, Higgsfield and Adobe Firefly out of the box. Any other provider can be added with a pricing template, no code.
Provider API keys are encrypted at rest. Each workspace is isolated, no workspace can see another's data. SpendGuard never forwards your SpendGuard key or attribution headers to upstream providers.
Nothing breaks. After 14 days of everything you drop to Free: tracking, caps, alerts and subscriptions for up to 3 people keep working. Pick Team or Agency whenever you want attribution, the Analyst or rebilling.
Set up SpendGuard in minutes and see your first attributed spend today.
Start your free trial