Recall's Free Tier in Practice: What 10 AI Summaries a Month Covers

What $0 actually includes

We read Recall's pricing page on 4 September 2026 and counted what the free plan gives away before it starts rationing anything. The capped item is the one everyone quotes, 10 AI summaries a month, but the uncapped items are the more interesting half of the page.

On Free, Recall gives unlimited read-later saves of articles, videos and PDFs, unlimited notes, a visual knowledge graph built from note links, and API and MCP access (recall.it/pricing, HTTP 200, read 4 September 2026). The last one deserves a pause. Programmatic access is the sort of thing most tools in this category hold back as a paid lever, and Recall hands it over at zero. A free-plan user can point Claude, Cursor or ChatGPT at their own library on day one.

That matters more than the summary count, because saving is the part of a reading habit that has to survive years. Recall does not meter it. Whatever else is true about the free tier, the durable layer underneath it is not the part being rationed.

One caveat on that MCP access, and it applies to every Recall tier rather than only the free one: their own developer docs say "The MCP server currently provides read-only access to your knowledge base. Write capabilities are coming soon" (docs.recall.it/developer/mcp, read 4 September 2026). An assistant can search and read your cards. It cannot put anything back. We have written about that asymmetry separately in can Claude save to Recall.

Ten summaries, spread over a month

Ten a month is one every three days. That is the number to hold your own habit against, and it is worth doing the arithmetic rather than guessing at it.

Workflows that fit comfortably: you save widely and summarise selectively. One long report a week, plus a book chapter and a conference talk, puts you at roughly eight. If your instinct is already to read most things yourself and reach for a summary only when the source is long or badly structured, ten is not a limit you will feel.

Workflows that do not fit: anything with a research burst in it. A conference week, a two-day sprint through an unfamiliar subject, a Sunday spent clearing a backlog of recorded talks. Those are the exact moments when a monthly cap binds, because the cap is monthly and your attention is not. Three summarised talks a week is twelve a month, and you are over before the last week of it.

There is a second-order effect worth naming. A per-action cap is indifferent to the size of the action. A four-hour lecture and a 600-word blog post each cost one of your ten. The rational response is to spend the allowance on your longest, densest items and read the short ones yourself, which is probably the better habit anyway. But it means the cap quietly pushes you towards one shape of use, and if your reading is mostly short pieces, ten summaries is worth less to you than the number suggests.

What the free tier does not include

This is where readers get caught, so it is worth being exact. AI chat over your knowledge base, quiz and spaced repetition, and the automatic knowledge graph are Plus features, not free ones (recall.it/pricing, read 4 September 2026). The free plan's graph is the one you build yourself out of note links; the automatic organisation is the paid one. Smart tags, Listen Mode text-to-speech, Augmented Browsing, multi-language, and bulk import of up to 1,000 bookmarks or 10,000 Markdown notes also sit on Plus.

The prices, from the same page on the same date: Plus is $10/month billed yearly, and the page's structured data also lists a monthly option at $12/month billed monthly. Max is $38/month billed yearly, with $48/month billed monthly in that same structured data. There is a 30-day refund and a 20% student discount. Max adds 1:1 onboarding, a choice of model, and bulk AI actions up to 100 items. Recall's own page describes the product as "Used by 700,000+ lifelong learners", which is their claim, on their site, read on that date.

So the practical shape of the free tier is this: save everything, organise it by hand, summarise ten things, and ask questions of the library from outside the app over read-only MCP rather than inside it.

A cap and a budget ration different things

A per-action cap and a metered token budget are not the same mechanism with different arithmetic. They fail in different places, and the failure mode is the thing you are actually choosing between.

A cap of ten is predictable. You always know what one summary costs, because it costs one, and you never have to model anything. The failure mode is that the eleventh interesting thing waits until next month regardless of how small it is, and the cap has no memory of how light your previous month was. It suits steady, uniform use and punishes bursts.

A budget is the opposite trade. It buys many cheap actions or a few expensive ones, so a small question costs a small amount instead of a whole unit. The failure mode is that you can spend it badly. Nothing stops you burning the month on low-value asking in the first week, and unless the running balance is visible you end up self-rationing against a number you cannot feel.

Neither model is the right one. If your use is uniform and infrequent, a cap is simpler and you will rarely notice it. If your actions vary a lot in size, or you work in bursts, a budget is more forgiving and hands you the responsibility for spending it well.

Where BrainTube stands

BrainTube's free plan is $0 with 30 BrainTokens a month (brain-tube.com/pricing). Saving is free and earns +1 BrainToken, so building the library is not the metered part. BrainTokens are spent on AI actions only, and an action costs the same on every plan, so upgrading does not make an action cheaper; it gives you more of them. MCP access and bt_ API keys are included on every plan, Free included, over one read-write server covering the whole library.

The honest other half. Thirty BrainTokens is a budget, not an unlimited anything, and a month of heavy asking will exhaust it. That is the budget failure mode described above, and BrainTube has it. Bulk YouTube import on the free plan is capped at 10 items. Proactive AI Insights is a real server-enforced tier gate at Pro and above rather than a metered feature. And there are no native iOS or Android apps, only add-to-home-screen; Recall's 19 August 2026 changelog entry records a mobile share sheet on iOS and Android, which is something we do not match.

We are not going to tell you which model is better, because the answer depends on whether your reading is steady or bursty. A side-by-side sits at braintube vs recall.

How these figures were checked

Recall's plan contents, prices, refund window, student discount and the "Used by 700,000+ lifelong learners" line were read from https://www.recall.it/pricing on 4 September 2026, HTTP 200. The $12 and $48 monthly figures come from that page's structured data on that date; the visible plan cards show the yearly-billed $10 and $38. The read-only MCP quotation was read from https://docs.recall.it/developer/mcp on 5 September 2026, HTTP 200; the separate API reference at docs.recall.it/developer/api, read the same day, is read-only too. The mobile share sheet comes from the 19 August 2026 entry on https://feedback.recall.it/changelog, read the same day. Nothing in this post rests on a figure that was not re-verified live on 4 September 2026.

Claims about BrainTube describe pricing and functionality running in production on 4 September 2026, confirmed against https://brain-tube.com/pricing on that date. If you spot something here that has gone stale, tell us and we will correct it with the date attached.