Free
For individuals trying memory in their own agent or editor$0
Free forever, no card
Start free- Unlimited retrievalSearch and briefings
- 500K tokens learned a monthAbout 500 conversations
- 100 written answers a monthFrom ask and context
Pricing
Unlimited retrieval. Pay for what Geniffy learns.
Learning is counted in tokens. A conversation is about 1,000, so on Pro it costs about $0.004 to learn one.
ProRecommended
For developers shipping an agent to real users$20/month
Try it free for 14 days, no card
Start 14-day free trialor buy Pro now- Unlimited retrievalSearch and briefings
- 5M tokens learned a monthAbout 5,000 conversations
- 2,500 written answers a monthFrom ask and context
- Extra learning, up to your capOnly past your plan
Max
For products serving thousands of users$100/month
or $1,000 a year
Get Max- Unlimited retrievalSearch and briefings
- 25M tokens learned a monthAbout 25,000 conversations
- 12,500 written answers a monthFrom ask and context
- Extra learning, up to your capOnly past your plan
- Priority support
Scale
For companies running agents in production$399/month
Invoiced monthly, no card
Talk to us- Unlimited retrievalSearch and briefings
- 100M tokens learned a monthAbout 100,000 conversations
- 50,000 written answers a monthFrom ask and context
- A shared Slack channel with us
- Our DPA, invoiced monthly
A larger company? For a security review, a data agreement or a price for your volume, talk to us directly.
How the price works
- 01
Save
FreeSend conversations, notes, files and pages. Every save is kept and searchable at once.
- 02
Learn
Counted in tokensGeniffy reads each save with a model and turns it into memories. Past your plan, $5 per 1M tokens.
- 03
Read
Never meteredBriefings and search run no model, so your agent can read its memory on every turn.
- 04
Answer
Counted in answersWhen Geniffy writes the answer for you, with ask or context. Past your plan, $5 per 1,000.
Estimate your month
A support chat is about 1,000 tokens, a four-minute call about 1,200.
Your plan
Pro + extra learning$45/month
- Tokens learned10M a month
- Learning a conversationabout $0.0045
- Briefings and searchUnlimited
Every plan
- The API, SDKs, MCP and framework packages
- Unlimited end users, each with a sealed memory
- Keys limited to one user, for their own app or device
- Answers cite their source, or say there is no record
- Saving never stops: past a plan, saves are kept and searchable
- Download or erase everything, any time, and never used to train any AI
Plan details
| What | Free | Pro | Max | Scale |
|---|---|---|---|---|
| AI apps that save | Two, and every app reads | Every one | Every one | Every one |
| Past your plan | Saves wait, kept and searchable, for next month | Saves wait for next month, or extra learning carries on, up to your cap | Saves wait for next month, or extra learning carries on, up to your cap | Saves wait for next month, or extra learning carries on, up to your cap |
| Support | Docs | Priority | A shared Slack channel |
Prices in US dollars.
Questions
Will my agent run out in a day?
No. The calls an agent makes on every turn, briefing and search, are never metered. A plan counts only what Geniffy learns from what you save, and the answers it writes for you. A coding agent saves about 25,000 tokens an hour, so Pro covers a full-time developer's month.
What happens when I reach my plan?
Nothing stops. New saves are kept and found by search at once, and learned when the month resets or your plan goes up. On Pro, Max and Scale, extra learning can carry on now, up to a monthly cap you set.
What is a token learned?
A token of the text Geniffy reads to make memories: what your app sends to POST /v1/memories, and the files and pages you add. A conversation of 1,000 tokens uses 1,000, and a ten-word note about 15. Saving, briefings and search are never counted.
What is a written answer?
One Geniffy writes for you with a model, from /v1/ask or /v1/context. Most agents put the briefing in their own model's prompt instead. Past your plan's answers, the same calls return the matching memories with their sources.
How does the free trial work?
Every account starts with 14 days of Pro: no card, nothing to cancel. When it ends you are on Free unless you choose a plan, and everything you saved stays.
Which plan does my app need?
Most apps start on Pro, and the estimate above picks a plan from your own numbers. Each of your users gets a sealed memory of their own, a space, and all of their use counts toward your plan. Busy apps use Max, and companies running agents in production use Scale.
Will the meters change?
No. A plan meters two things, tokens learned and written answers. A new way in, an AI app over MCP or an app you connect, counts as learning in the same plans.
Can I cancel any time?
Yes, in one click. You keep your plan to the end of what you paid for, then move to Free. Your API keys can still read your memory, and you can export it any time.
Do you offer refunds?
Yes. Ask within 7 days of your first payment and you get it all back.
Who can see my memory?
Only you and your API keys. It is never used to train any AI model. Export it or erase it whenever you like.