Pricing that shows its work.
Three tiers. Every query is metered at model cost plus 30 percent and shows what it cost on the answer it paid for. Spending is prepaid; nothing can bill past credit you bought.
Annual billing is two months free, shown in each price below.
Free
For finding out whether this works on your questions.
- 25 papers in your library
- 10 corpus questions a month
- 3 full synthesis runs, total
- The included 13,900-paper searchable corpus
- No card required
The synthesis runs are the real product. Three is enough to judge the output on a topic you know.
Pro
For one person doing real research.
- 1,000 papers in your library
- No cap on questions or synthesis runs. Credit-based: about 375 runs a month on the included $15.
- Per-query cost and grounding shown on every answer
- Cost preview before every heavy run; nothing spends past your balance
- MCP access at launch: your agents can search the corpus and ask grounded questions
Team
For a desk.
- 10,000 papers in your library
- Five seats
- $60 of model credit included monthly
- Everything in Pro, on every seat
Set up per desk, by contact.
Measured, per tool.
| tool | time | model cost |
|---|---|---|
| Corpus question | ~30 s | ~$0.05 |
| Disagreement map | ~25 s | ~$0.04 |
| Research program | ~60 s | ~$0.09 |
| Strategies with adversarial review | 3 to 4 min | ~$0.18 |
Medians from the measured capture session of 2026-06-11, against the real corpus on production models. The metered price is model cost plus 30 percent.
The only rate in the product.
Every query is metered at model cost plus 30 percent, itemized, and shown on the answer it paid for. That is the only rate. The tiers decide how much credit is included.
Pro includes $15 of metered credit each month; Team includes $60. At the measured deep-tool cost, the $15 is about 375 runs a month.
Spending is prepaid. Every heavy run shows its estimated cost before it starts, and a run your balance cannot cover does not start. You get a notice when credit runs low. There is no overage and no mechanism to bill past credit you already bought.
If a month runs heavy, a booster adds $25 of credit for $25, paid up front, applied immediately.
Included credit does not roll over. The meter resets with your billing month.
- What counts as a synthesis run?
- Any of the heavy tools: a research program, strategies with the adversarial review, a combiner or sweep run. A corpus question is lighter and counted separately on Free. On paid tiers neither has a per-run cap; the bound is your credit balance.
- What happens when my credit runs out?
- Paid queries stop, with the numbers shown, and work in progress completes. A booster tops the balance up. Nothing can bill past credit you already bought. Automatic top-up at a threshold you set is on the roadmap.
- What if a run fails, or comes back saying the evidence is not there?
- A technical failure never consumes a run and never bills you. On the Free tier, a run that ends in an honest bail does not count against your three. The three runs exist so you can judge the synthesis on questions you know.
- Which model does it use?
- Frontier Anthropic models routed through OpenRouter. Each answer shows the model cost it incurred.
- Can I bring my own API key?
- Not yet. It is on the roadmap. All usage runs on included and metered credit.
- How do I get the Team tier?
- Write to [email protected] and we set your desk up directly. Billing starts when your seats do.
- Can I cancel or change plans myself?
- Yes, from the billing portal, immediately. Annual plans are refunded pro rata where the law requires it and case by case otherwise; write to us.
- Do you train on my library or my queries?
- No. Your uploads and your questions are yours.
- Is there a trial?
- The Free tier is the trial and it does not expire. Three full synthesis runs is enough to judge the output on your own questions.
No card. If the output is not worth $39 to you, the Free tier already told you so.