Recall 2.0 Turns Your Knowledge Base Into a Personal AI

Recall’s core pitch is four words: “The edge is your knowledge,” the team said. Whether you believe it depends on how seriously you take the premise that generic AI has made most information advantages disappear.

I think the premise is mostly right.

Recall 2.0 landed this month. It’s a meaningful departure from what version 1.0 was, and the gap between the two products is worth spelling out in full before we get to the numbers.

The original Recall 1.0 was a knowledge-base app. You saved articles, notes, research clippings. It organized them, summarized them, drew connections between them. Useful, genuinely, but the category was crowded. Notion does something adjacent. Obsidian has a devoted user base that treats it like a religion. What 1.0 didn’t do was let you talk back at your archive. You were still the retrieval layer. You still had to remember that you’d saved something before you could find it.

Version 2.0 changes that. Completely.

The product now runs a conversational AI interface that’s grounded in your saved content rather than a generic training corpus. The distinction matters more than it sounds. You’re not dumping documents into a chat window each session and hoping the context window holds, the way people were doing in 2023 when the workflow hacks were still being figured out. You build the archive over time, and the AI learns the shape of what’s in it. Then you query it. Not “search” it, exactly. Query it, the way you’d ask a research assistant who’s read everything you’ve read in the last three years.

The framing on their Product Hunt page is clean: “Talk to your knowledge, the internet, or both.”

That toggle, knowledge-only versus open internet versus both, is the sharpest product decision in version 2.0. Most AI products don’t give you that granularity. You’re either querying a live model that doesn’t know what you’ve personally saved, or you’re stuck inside a walled archive that can’t cross-reference new information. Recall’s approach lets you decide scope at the prompt level, which is genuinely different.

The sample prompts the company uses aren’t the usual demo theater. “Find the exact clip in my podcast.” “Condense my research on X.” “Compare this new study to the three I saved last quarter.” Those are the kinds of requests that would break most general AI tools because they require knowing what’s in a specific person’s archive, not just what’s on the internet or in a training set.

Now the numbers. Recall claims 500,000 professionals on the platform. Logos on the site include Stanford University, Bloomberg, NYU, and LinkedIn. I can’t verify whether those represent formal institutional agreements or individual users who happen to work at those organizations. That distinction matters, and it’s a common sleight of hand in B2C-to-enterprise marketing. So take the social proof at a slight discount. Still, 500,000 users is a number that takes the product out of the “promising experiment” category if it holds up.

The Product Hunt performance deserves its own sentence. The launch hit 275675% on whatever growth metric they’re tracking there, which is the kind of number that reflects genuine distribution momentum rather than quiet organic growth.

There’s a substantive body of research behind why this product category keeps attracting serious attention. The PKM research community at NYU’s information science department has been examining how knowledge workers manage personal information since well before AI made it commercially interesting. The core finding across 6 or 7 years of that work, going back roughly to 2020, is that retrieval failure isn’t the problem people think it is. The problem is context collapse. You save something, you remember saving it, but you can’t reconstruct why it seemed relevant at the time or how it connects to what you’re working on now. A system that maintains relational context between saved items, and lets you interrogate that context conversationally, solves a different and harder problem than basic retrieval.

That’s what Recall 2.0 is actually building toward. Whether it gets there is a different question.

The technical additions in 2.0 beyond the conversational layer: API access and MCP support. Those two items together signal that the company isn’t positioning Recall purely as a standalone consumer app. API access means you can pipe your personal knowledge base into other workflows. MCP support means it can talk to other tools in a more structured way. For knowledge workers who’ve built their own systems across Notion, Obsidian, or something custom, this is the part that makes Recall composable rather than isolated.

Worth noting what hasn’t changed and what that means. The fundamental model is still: you bring the content, Recall provides the architecture and the AI layer on top. That means the product’s value scales directly with how disciplined you are about saving material. For someone who’s been archiving consistently since 2020, this could be genuinely powerful. For someone starting fresh, there’s a cold-start problem that no amount of good AI design fully eliminates. The archive has to exist before the intelligence can work on it.

There are 29 direct competitors in the second-brain and PKM space that I can count from the last 18 months alone. Most of them are dead or stalled. The ones still standing tend to have either a strong community, like Obsidian, or a strong enterprise sales motion, like Notion. Recall’s bet is that the AI layer is the differentiator, which is a reasonable bet in 2026 and would have been a losing one in 2020 when the models weren’t good enough to make it work.

The 1,500 or so serious PKM enthusiasts who represent the core of the early adopter market for products like this don’t move the revenue needle by themselves. Recall needs the 500,000 number to be real and to be growing. The product has to work for people who won’t obsessively maintain their archive, which is most people, not just the 30 or 200 power users who would optimize any system.

I don’t know whether Recall 2.0 clears that bar. The product is genuinely interesting and the 2.0 upgrade is a real step up from what 1.0 was. The toggle between personal knowledge and open internet is clever. The sample prompts show actual use cases rather than marketing abstractions. But the user-discipline problem is real, and it’s not solved by better AI. It’s solved by product design that makes saving things effortless enough that people actually do it consistently. That’s a harder UX problem than building the conversational layer.

The team’s own framing is probably the right north star for evaluating the product over time. “The edge is your knowledge,” the team said. If that’s true, the product only works for people who’ve built an edge worth having.