Plans and usage
Free to read, pay to learn. Saving, briefings and search are never counted. A plan sets how much Geniffy learns from what you save each month, said in hours of AI work, and the answers it writes.
The plans
| Free | Pro | Max | Enterprise | |
|---|---|---|---|---|
| Price | $0 | $20 a month, or $200 a year | $100 a month, or $1,000 a year | Talk to us |
| Learning a month | 10 hours of AI work (250K tokens) | 80 hours (2M tokens), 8x Free | 400 hours (10M tokens), 5x Pro | Set for you |
| Written answers a month | 50 | 1,000 | 5,000 | Set for you |
| AI apps that save | Two (every app reads) | Every one | Every one | Every one |
| Extra learning | No | Yes | Yes | Yes |
Every plan has the API, the SDKs, MCP and the framework packages, unlimited end users with a sealed memory each, and unlimited saving, briefings and search. Every account starts with 14 days of Pro, no card needed.
What is counted
- Learning. Geniffy reads what is saved and learns from it: the facts, where each project stands, the promises, what happened and the lessons. An hour of AI work is about 25,000 tokens: an hour of a coding agent, its tool steps included, or about 25 long chats. Each save counts the tokens read to learn it, at least 100; a source sent again counts only what is new in it, and text sent again unchanged counts nothing.
- Written answers. An
ask()that comes back with an answer, or acontext()block. A question nothing supports, or one past your answers, is free.
Everything your app's users save and ask counts toward your plan, through every door: the web app, MCP, the plugins and your keys.
Past your plan
Nothing is lost and nothing errors:
- A save is kept and found by search at once, and learned when the month resets, oldest first.
- A question gets the matching memories with their sources, without a written answer.
- With extra learning on (Pro and Max), saves are learned and questions answered as before, at $15 per million tokens and $15 per 1,000 answers past the plan, billed with your next renewal and never past the monthly cap you set in Plan and usage. A plan paid by UPI can't take a charge after the fact, so extra learning needs a card.
Only when what waits reaches ten months of your plan's learning is a save refused, with 402 allowance_used
(Errors).
Reading your use
Every response carries this month's figures, so your app can watch its use without asking:
| Header | What it says |
|---|---|
X-Geniffy-Usage-Learned |
Tokens learned this month |
X-Geniffy-Usage-Included |
Tokens your plan learns a month |
X-Geniffy-Usage-Resets |
When the month resets |
GET /v1/usage spells it all out: tokens learned and waiting, answers, extra learning and its cap,
and where you stand against the plan (ok, near at 80%, over, extra, or paused). In the SDKs it is
usage().