PharmaDB
Pricing · lookups capped at $1 per run · AI at 20% over Anthropic v.05.2026 · current rate card
Pricing · pay for what you use

Pay for runs, not for seats.

You pay for two things: graph lookups, and AI usage. Lookups are capped at $1.00 per run, no matter how many we make. AI is billed at 20% over Anthropic's published rates. Both line items show in your chat as the AI agent works. No seat charges.

A run is
One question to the AI agent. You ask, the AI agent answers. That whole exchange is one run.
A lookup is
One query the AI agent fires against the graph while answering. A typical run uses a handful.
Pay as you go
For pilots and one-off questions
$0.20/ run, starting band
Top up and start
AI usage · model rate card

What you pay for the AI itself.

Anthropic publishes a rate per million tokens for each Claude model. We add 20% on top to cover serving, retries, and tool orchestration. The model that handled your chat shows up in the chat, so you can always see which rate applied.

Model
Input · $ / MTok
Output · $ / MTok
You pay · in / out
Claude Opus 4.7The AI agent's default model
$5.00
$25.00
$6.00 / $30.00
Claude Sonnet 4.6Faster, lighter queries
$3.00
$15.00
$3.60 / $18.00
Claude Haiku 4.5Quickest, cheapest, retrieval-heavy work
$1.00
$5.00
$1.20 / $6.00
Rates as of 2026-05-22. If Anthropic changes their published prices, ours move with them.

Questions, answered.

What's the difference between a run and a lookup?

A run is one question to the AI agent. A lookup is one query the AI agent fires against the graph while answering. A single run usually uses a handful of lookups, and the lookup charge is capped at $1.00 per run no matter how many. So $1 is the most you'll ever pay for lookups in a single the AI agent answer, period.

What's the most a single run can cost?

$1.00 for lookups, plus the AI usage for whichever model handled it. A simple Opus question with a few thousand tokens is usually under $0.50 total. A heavy multi-tool Opus question that hits the 50+ lookup band tops out at $1.00 in lookups plus the model's token cost on top.

Can I switch between PAYG and Pro mid-year?

Yes. Any unused PAYG credits move into your Pro pool at full value. It also works the other way: if you don't end up using a Pro year, the balance becomes PAYG credit. No double-charging, no re-signing.

Why charge 20% on top of Anthropic's rates?

It covers running the service: hosting, retries when calls fail, and the tool orchestration that lets the AI agent call the graph alongside the model. We show it as a separate line so you can see what each query actually cost.

Do credits really expire after 12 months?

Yes, from the date you topped up. If you're on Pro, each renewal refreshes the $5,000 credit pool for another year. We do this to keep the rate card simple, not to clip credits.

What's not included at any tier?

Named buyer-to-seller transaction listings. The trade data is aggregated and anonymised at the counterparty level. We show lane totals, country-pair flows, percentiles, and supplier relationship graphs. We never show "company X bought N kg from company Y."

Want to try it first?

Book a 30-minute call. Bring three sourcing questions. We'll run them on the live data with you. No top-up needed, and you keep whatever we find.

Book a call