Pricing
The free tier, in writing
Everything live today is free with attribution — and stays that way. Paid tiers add guarantees, not gates.
Free
Live todayEverything you can touch on this site right now. Build with it, ship with it, let your agent call it.
- ✓All 27 verified models — REST API and MCP server
- ✓Full specs, contracts, and the public changelog
- ✓Embeddable calculators (one script tag)
- ✓Fair use: 5,000 model runs a month per app — a guideline, not a meter
- ✓Attribution: keep the badge on embeds, credit API results
Pro
Early accessFor products that put our numbers in front of their customers and need promises, not just uptime.
- ✓SLA and support with a human on the other end
- ✓Attribution-free embeds and API results
- ✓Volume headroom and burst room
- ✓Spec-version pinning — your integration never moves under you
- ✓Private usage dashboards
Enterprise
By designFor platforms that want the engine as their own — or want models that don't exist yet, built and verified the same way.
- ✓White-label calculators and planning flows
- ✓Custom models: spec'd, dual-implementation verified, maintained
- ✓Verified eval datasets for financial AI testing
- ✓Facts-registry feed — tax-year changes, sourced, the day they land
- ✓Contracts, DPAs, compliance documentation
The questions worth asking
Will the free tier stay free?
Yes — that's the point of writing it down before anyone asks. The models, the API, the MCP server, and embeds stay free with attribution. Paid tiers exist to add guarantees (SLAs, pinning, white-labeling), never to take back what's free today. Fair use is a published guideline (see below), and if enforcement ever becomes necessary it arrives with notice and generous headroom — not retroactively.
What's fair use, in numbers?
5,000 model runs a month per application. It's a guideline, not a meter — nothing blocks you at 5,001. If you're consistently above it, you're succeeding, and that's a Pro conversation, not a shutoff. The only hard limit is a burst backstop (120 requests a minute per IP) that exists to stop runaway scripts, never real users.
What counts as attribution?
Embeds carry it automatically — just don't strip the badge. For API and MCP results shown to end users, a visible “Powered by Worthune” with a link does it. Internal tools and evaluation use need nothing.
Why no published prices on Pro and Enterprise?
Because we'd rather set them with the first design partners than guess at them in public. Early partners get terms that reflect that. Once pricing settles, it goes on this page — and nothing that was free moves behind it.
Do you track my usage?
Only in aggregate — daily counters per model and tool, no IPs, no user agents, no inputs. We publish the same numbers we see, at /api/v1/telemetry.