Querri
Question Anything
680 followers
Question Anything
680 followers
Querri brings your company's data together across 100+ platforms, giving every department a shared foundation for reliable, repeatable answers. Connect, clean, and combine your data, then teach Querri the definitions and context that matter to your business. Ask questions in plain English, build dashboards and presentations, automate recurring analysis, or work with your data directly from Claude and ChatGPT.
This is the 3rd launch from Querri. View more
The Library
Launching today
The Querri Library is where your data gets organized around the questions your business asks, so every answer comes from the same trusted, shared foundation.. The library learns your definitions, business context, and past analysis, so its answers are grounded in how your business actually works. And when it doesn’t know, it says so, instead of making things up.
Free Options
Launch Team
Hi team, congrats on the launch. Isn't it same thing those big data warehouse players are building? like Cortex from snowflake and Geni from databricks? and I think the challenge for the data query using nlp is governance and how to make sure you answer is correct
@gavin_fgg0815 Hey Gavin, thanks! Yes, getting the answer right is the most important challenge.
Genie and Cortex are the agent layers from SF and DB. These are their agent building tools and somewhat similar. What sets Querri apart is that it's the full stack, from ingestion to data lake to semantic layer to chat and dashboards, all in one package with one bill that can be set up in a few minutes and managed by the whole team. We're really aiming to support the teams that don't have the time, budget and team to do multi-quarter data projects.
As to accuracy, we take that super seriously. All the answers are structured, so you can trace the number in plain english (and SQL as needed) all the way back to the source data. We can only do this because we've got the raw data all the way through the computations, and because the core of the agent is built around repeatable analyses not just spitting out an answer which can and will sometimes be hallucinated.
@daveaingram Hi Dave! Thanks for the answer, still it is very hard to verify your number is right, unless those are pre-orchestrated (but pre orchestration ruin all the agents autonomous). Before, I am struggling to show this to top leader, because I don't have 100% confidence, and it will be very bad to report them a wrong number. When you have a team do this in dashboard, at least you have people to check multiple times
@gavin_fgg0815 I think this is the big challenge of the AI era ... fast answers that look right but don't get verified. The good news is, with Querri there's literally nothing you can't deeply inspect both in plain English and in the code. So there's nothing you could check in a dashboard or an Excel formula that you couldn't walk back through here, and we lay it out a lot cleaner imho.
The other thing we can do with AI and do extensively in Querri is double / triple check the numbers and very deliberately surface caveats and alternate readings right in the chat. In our evals, we're 6x more likely to show you an alternate reading that Claude Code does without going and explicitly asking it for alternate readings (which pretty close to nobody does ... and they rarely do with dashboards or other tools either).
That said, pre orchestration is definitely great when it can be done, which is what we do in the Library. You can define exact SQL for a KPI, and build precomputed Views for common questions, but unlike some systems that only allow you to then read from those, Querri can fall back to the raw data (if you've given access to it...all configurable) and still get the autonomous agent capability.
Hi Product Hunt! Dave here, founder and CEO of Querri.
Today we’re launching the Querri Library and the Librarian, the biggest release in Querri’s history. Querri is the question-first data platform: you start with the questions your business asks, and Querri builds what’s needed to answer them, then keeps learning your business.
The problem → solution
AI made asking your data a question easy. It didn’t make the answer yours. Point a general-purpose AI at real company data, with hundreds of tables and five definitions of “active customer,” and it picks one. Ask again tomorrow and it may pick another.
So we stopped trying to make the AI a better guesser and changed what it’s standing on. The Library is a map of your business: the questions your team keeps asking, the KPIs that answer them, the definitions behind those numbers, and the data they’re computed from. The Librarian is the agent that builds it with you, in plain English, and answers from it.
What makes it different?
A chatbot on top of your data is easy to build. I can say that with some authority, because that’s roughly where Querri started. What’s hard is making the answers right, and yours. Every answer in Querri shows the rows, the calculation and the sources it read. Every piece of data loaded gets 180+ quality checks. When a question has more than one reasonable reading, Querri tells you. In our benchmarking, Querri flagged the other defensible answer on 73% of its responses, versus 12% for a leading general-purpose AI agent given the same data. And when the data is too messy or isn’t there, it tells you what’s missing instead of handing you a confident number that happens to be wrong.
And all of this feeds into chat, dashboards, MCP for Claude, ChatGPT and the like, and APIs and CLI for further extension.
Key features & use cases:
Start from your questions before you connect a single file
125+ connectors, plus spreadsheets and files
KPIs defined once and tied to Revenue, Cost, Efficiency, Quality or Risk
Business rules you settle in chat (“sales means paid orders only”) saved as facts
Paste a screenshot or sketch of an old dashboard and it rebuilds it, calculations included
Monitors that watch a KPI by region, store or rep and tell you which one moved (set up with an agent but run without every time the data updates)
Share a cleaned view with another team instead of the raw tables via access controlled workspaces .. we’re talking real life data work here after all.
While you sleep, Querri “dreams” about your business and suggests new questions and KPIs. Nothing changes without your OK.
Who is it for?
Ops, customer success and revenue teams whose answers live across several systems. Small business owners doing the data work themselves. Consultants who want each client’s definitions to live somewhere. Data teams who want give the business an easy, accurate and reliable front door and free their issue queue. And software companies that want to white-label it for their own customers.
It learns your business, and it never trains a model on your data. We’re SOC 2 Type II, HIPAA and ISO 27001:2022 certified.
My ask: bring the question that tripped up the last AI you tried, and tell us what Querri does with it. I’ll be here all day reading and answering every comment.
The full story is on our blog: https://querri.com/blog/introducing-the-querri-library-and-librarian/
Dave
P.S., want to really put it through the paces? Sign up for a trial with HUNTER1000 promo code and you'll get a 1000 bonus credits to get some serious value out of your data.
Building Querri has been quite a journey!
Our mission, from the very beginning, has been to make working with data accessible to everyone. In a world where businesses collect more data than ever, understanding that data is no longer just a nice-to-have. It can make or break important business decisions.
But getting data together from dozens of different systems is hard enough. Getting AI to give you reliable answers based on your business's definitions of revenue, churn, or other important metrics is another challenge entirely.
We've spent thousands of hours tackling this problem. Because it matters.
AI is an incredibly powerful tool, but it needs to understand your business context. We're also huge believers in keeping humans in the loop. Data has so much nuance, and it would be naive to assume AI automatically knows how your business defines everything or how you want to handle every inconsistency.
That's why we've built so many ways for people to guide the AI, teach it their business context, and stay in control of how their data is cleaned, analyzed, and understood.
And security isn't an afterthought. Querri has completed a SOC 2 Type II audit, is ISO 27001:2022 certified, supports HIPAA compliance, and never uses customer data to train AI models.
I'm incredibly proud of our team and what we've built together.
We're excited to share it with the Product Hunt community, and we'd genuinely love your feedback, questions, and ideas. That's how we learn, grow, and keep getting better!
congrats team! rebuilding an old dashboard from just a screenshot sounds kinda wild what happens if the numbers don't match the original, does it tell you why?
@suryansh_tiwari2 Library doesn't just compute, it understands your business context, hence it is highly likely that it will point out the discrepancy right away
@suryansh_tiwari2 thanks!
And yes, we're really deep in "agent loops" over here. One builds the design of the dashboard and puts placeholders for the data, while another goes into The Library and looks at the data you've loaded and any business definitions you've created already. It then creates what we call "projects" for the different analyses (which can later be automated without the agent so the dashboard is live and trustable) and then it all comes together in the finished dashboard. All of those agent chats are visible, so anything it's struggling with like missing data or data quality issues can all be handled conversationally.
How do you handle when the underlying data changes, business definitions evolve or 2 equally valid interpretations exist? Who ultimately owns the truth; the data, semantic layer or the person asking ques?
@manjunath19 these are all very real world scenarios.
Ultimately, we see the choice going to the "Data Owner", who is the administrator of a particular library or workspace. When a consumer is just asking questions, the question is going to filter through the semantic layer first and through that to the data.
If there are clear definitions in the semantic layer then the user will get a quick, direct answer. If the question is ambiguous, the semantic layer not set up yet, or no definitions exist, then the agent knows it needs to read raw data and work extra hard. In this case we do several tasks in parallel and the agent deliberately looks for multiple readings and surfaces both its best guess to the user as well as the other reading and the caveats that are found in the data.
Then, the librarian runs overnight and can check for complex or ambiguous questions that were asked and surfaces those to the data owner along with suggestions on how to update the semantic layer, so the next time someone asks a similar question, the answer can be settled.
There are also a lot of things that can be caught and surfaced even earlier when a user is setting up Querri. We run 180 or so checks on every column of data that comes in looking for everything from gaps in time series data to hidden characters to mixed formatting. These get surfaced automatically to the agent until they're resolved, but also on a Data Audit screen so the owner can look through them and resolve during setup.
All of this gets refreshed and new issues automatically surfaced to the Data Owner after data updates.
Dave and the team have been amazing to work with as I've looked to roll out better visibility across a number of systems across multiple companies in our portfolio. There has been no system they've been unwilling to try and build connectors for which has been a huge time benefit for me in a job where time is a premium.
@neil_amin Thank you so much! It means a lot to hear this. Our team is always willing to tackle challenges and find solutions, even when they're not straightforward. We know your time is valuable, and we're excited about the opportunity to make your job a little easier. Really appreciate your trust and support!
@neil_amin Thanks Neil - you and your team have been a pleasure to work with. It's been great enabling you and your firm to see data across many systems and also being able to give consistent views of data for each of your portfolio companies.
Congrats Dave, Amy and the Querri team.
The point about the same question getting a different answer tomorrow is the real problem with AI on top of data.
When two teams define "active customer" differently, does the Library force one definition, or can both live side by side?
@genedai Thank you, Gene! And that's a great question.
Both definitions can be valid! What Sales considers an "active customer" might differ significantly from how Customer Success defines it.
The goal isn't to force everyone to agree on one definition if there are legitimate differences. It's to make those differences explicit, with clearly defined metrics, so the AI isn't guessing which one to use.
Ultimately, we want people to understand not just what the number is, but how it was calculated and what it actually means.
That's exactly the kind of nuance we've been working so hard to address!
@genedai great question! ... Querri has "workspaces", and those are connected, so if you have a genuine need to track two different definitions you can put the global one in the shared workspace and override it in another. But inside of one workspace, our "Librarian" agent always looks at what already exists before creating new KPIs, business rules, Views etc, so if you try to create a duplicate rule it's going to ask you if you want to update or explicitly create something with a new definition. And all of this stuff can have "facts" hanging off of it, so you can clearly tell the AI about each definition and then when you ask about "active customers" going forward it would prompt you with something like "did you mean 'active customers' or 'active paying customers'?" .. or whatever the distinction is.