Same question, asked twice. Your AI isn't wrong. It's average.
WarpSearch sessions, agents, files…
❯Should we ask for a credit card at signup?
Thinking…
It depends on your business. Requiring a card: far fewer people sign up, but a much larger share of those who do end up paying.
Thinking.
✻ Claude Code
❯
➜ demo-projectmain · Claude Code
Terminal~/demo-project
AI alone
Generic. No source.
VS
+
WarpSearch sessions, agents, files…
❯Should we ask for a credit card at signup?
Trove LibrarianFinding the right book…
No, by default. Of 5,000 serious visitors, a card upfront gave 30 paying customers. No card gave 60. No card plus a focus on active users gave 100.
Lean Analyticspp. 220-221
Step 1: Ask.
✻ Claude Code · Trove connected
❯
➜ demo-projectmain · Claude Code
Terminal~/demo-project
AI + Trove
Specific. Lean Analytics, pp. 220-221
01 / 09
Real chats, October 2026. Answers shortened, meaning kept.
What you learn, nobody has to repeat.
The gotcha you find at 1am shouldn't die with the chat. Approve it once, and your next session already knows. So does your team's.
How ≠ Why
SkillsThe how
Reusable steps, written down on purpose.
The why
What's in your head stays there. Unless it's in Trove.
Don't just get an answer. Know why you should trust it.
Every answer names its source, down to the page. No black-box guesses.
Finds what you didn't ask for
Surfaces what you didn't know to look for, and says so when the right book isn't there yet.
Always current
Knowledge stays current automatically. Never a stale snapshot.
Disagreement, shown
When sources disagree, you see both, each with its own citation.
Looks familiar. Works differently.
Trove gives the agents you use the same expert books, with the source cited, and keeps what it learns.
Trove does all four
Expert books, with sources cited
One library, every agent
Finds what you didn't ask
Learns as it works
1 of 8 · Built-in memory
CMP · 000.1 · MEMOClaude, Codex, Grok
Built-in memory
It remembers you, inside one app.
Good at
Remembering your preferences in the one assistant you always use.
Falls short
Most people use a few agents, and what one learns stays locked inside it. Nothing it holds is a book you can cite.
Trove
One library that Claude, Codex, Cursor, Grok and GitHub Copilot all read and write. What one agent learns, the others know instantly.
CMP · 001.4 · NBLMResearch notebook
NotebookLM
Great for you to read. Closed to your agent.
Good at
Researching your own sources, when you're the one reading.
Falls short
Your agent can't open your notebook, so you copy answers across by hand.
Trove
Your agent opens the book itself, reads the chapter it needs and cites the page.
CMP · 002.1 · OBSDPersonal notes
Obsidian
Easy to write. Hard for your agent to find.
Good at
A private, offline place for your own thinking.
Falls short
Your agent searches by keyword, you file every note yourself, and sharing means handing over the whole vault.
Trove
Your agent finds the chapter from the table of contents, files what it learned, and you share one book with anyone.
CMP · 003.7 · NOTNNotes and wikis
Notion and wikis
A database, not a method.
Good at
Giving people one place to write things down.
Falls short
Your agent reads whole pages to find one answer, has no place to file what it learned, and nobody sees what agents read or couldn't find.
Trove
The librarian picks the chapter, your agent reads only those pages and cites them, manuscripts keep what it learns, and you see what agents looked for and didn't find.
CMP · 008.0 · BRAINEnterprise search, like Glean
Company brains
Finds everything. Vouches for nothing.
Good at
Searching everything a large organization already wrote.
Falls short
It indexes the drafts and old decisions too. Marking which one is true is manual, one document at a time.
Trove
The part a person vouched for, cited by page. A company brain can read it too.
CMP · 006.3 · GHUBCode and docs host
GitHub
Saving one lesson shouldn't need a pull request.
Good at
Knowledge about one repo, read by people who already work there.
Falls short
Every lesson is a commit waiting on a reviewer, everyone must pull to stay current, and your CEO or marketing lead doesn't work in GitHub.
Trove
Your agent writes what it learned straight into the book, every reader is always on the latest, and you share it with anyone, engineers or not.
CMP · 007.5 · SKLSAgent skills
Skills marketplaces
A skill knows the steps. Not why they're the steps.
Good at
Procedures your agent should run the same way every time.
Falls short
The reasoning behind a procedure lives nowhere your agent can read.
Trove
Books hold the why: decisions, trade-offs and exceptions, right next to your skills.
CMP · 005.9 · RAGBuild it yourself
DIY retrieval
Three chunks, no chapter.
Good at
Data that can't leave your network, when you have people to run it.
Falls short
You chunk, embed and re-index every time something changes, and your agent gets fragments.
Trove
No pipeline. Your agent reads whole books and cites the page.
Logos belong to their owners and name the products compared.
Don't take our word for it. Take theirs.
From people who use Trove.
Matan MarudiAI Engineer, Cyera
Trove gives me confidence my agents have the right context, grounded in trusted books instead of stale model knowledge or noisy web results.
New York
Hadar BarakCo-founder, Vest
As a founder I can't afford confident wrong answers. With Trove my agent shows me the exact page it's relying on.
Tel Aviv
Tom GranotFounder, Kintsu.ai, 300k+ YouTube subscribers
When my Codex builds, it almost never asks me about architecture. With Trove, it's like having the finest minds in the world review my code.
Amsterdam
Dani KamanovskyEngineering Leader for Security and AI
I was productive in week one, not month one.
Tel Aviv
Yotam FrommSenior Software Engineer, Sefaria
Trove is the first and last step in every field I work in.
Brooklyn, NY
Idan BenaunCo-founder & CTO, sled.ai
Trove has let me take my agentic workflows from 80% to 95% in a single Claude Code session.
Washington, DC
Matan MarudiAI Engineer, Cyera
Trove gives me confidence my agents have the right context, grounded in trusted books instead of stale model knowledge or noisy web results.
New York
Hadar BarakCo-founder, Vest
As a founder I can't afford confident wrong answers. With Trove my agent shows me the exact page it's relying on.
Tel Aviv
Tom GranotFounder, Kintsu.ai, 300k+ YouTube subscribers
When my Codex builds, it almost never asks me about architecture. With Trove, it's like having the finest minds in the world review my code.
Amsterdam
Dani KamanovskyEngineering Leader for Security and AI
I was productive in week one, not month one.
Tel Aviv
Yotam FrommSenior Software Engineer, Sefaria
Trove is the first and last step in every field I work in.
Brooklyn, NY
Idan BenaunCo-founder & CTO, sled.ai
Trove has let me take my agentic workflows from 80% to 95% in a single Claude Code session.
An "item" is anything in your library: a published book, your own document or a Manuscript.
An "agent read" is counted each time an agent opens a page.
"Publish" means putting a book or manuscript into the marketplace for others to use.
Your knowledge stays under your control.
Never used to train models
We don't train models on your content. Not for you. Not for anyone else.
Delete a book, it's gone
Removed from storage, not just hidden.
Delete your account, your library goes too
Nothing lingers behind.
No exceptions, not even us
Everything we know lives in Trove too, under the same rules as yours.
Private by default
Visible to you, and anyone you choose to publish to.
Questions, answered
The important things to know before you get started
A library of expert books and your own docs that your AI reads. Every answer shows the source it came from.
No. Claude Code, Codex, Cursor, Grok, and GitHub Copilot are where you work. Trove is the library they read from, so you keep using the tools you already have, just backed by knowledge they can cite.
No. The Trove librarian, your agent's own research assistant, finds the relevant books, reads them, and cites what it used. If nothing in your library fits, it says so and suggests the book to add.
Every answer comes back with the actual source and page cited, so you can check it yourself instead of taking it on faith.
A lot of what you actually need isn't on the web at all: your team's decisions, the paid books, the playbook your best engineer carries in her head. What is on the web is noisy and increasingly AI writing about AI. With a library, the shelf is stocked in advance and your agent reads books you chose, and it can cite them.
You see both, each with its own citation, instead of one confident-sounding blend.
No. Personal is free and built for exactly that, one person, one library.
No. We don't train models on your content, and we don't use it to train anyone else's.
No card for Personal, the free plan. Cancel Pro or Team any time and keep your plan until the end of the month or year you paid for. On a yearly plan, ask within 14 days of your first payment and we'll refund it in full. Monthly payments and renewals are not refunded.
Your AI is already powerful. Give it something worth knowing.
Your AI works from expert books and your team's own docs, and every answer shows where it came from.
The why
Popular
