Launch HN: Prized (YC S26) – Let non-engineer staff build secure internal tools

Posted by marinoseliades 4 days ago

Counter77Comment59OpenOriginal

Hi HN, we're Marinos and Hudson, founders of Prized (https://prized.dev)! Prized lets non-engineer employees describe the internal tool they need and get a full-stack app, wired to their company’s data and deployed behind the company’s sign-in, without them ever juggling API keys or connectors.

Here's a demo: https://www.youtube.com/watch?v=730MuYOfZTY

The way Prized provides security is by limiting what the agent can reach at the network layer and by keeping credentials out of the sandbox entirely. The sandbox never holds any keys or connector secrets, it only uses scoped session tokens that are stored as opaque placeholders. The real values are swapped into the request headers on our egress proxy. When production data is connected, the sandbox's network policy is set to deny by default so the only path out is via the proxy. Any call the agent makes to an external connector is reviewed by an LLM judge to prevent dangerous operations.

Prized is meant for the internal workflows that start as notebooks or spreadsheets but never become real tools because engineering has more important things to work on. One customer’s data scientist pasted in his personal fraud-detection notebook with hardcoded thresholds and all. After a few prompts, it became a published risk console connected to the company’s data with those thresholds turned into UI controls. Earlier today, we got off a call with them and most of their company is using it.

To do this, you need to give people the freedom to build without having unaudited access to company systems. We allow admins to scope data to specific users or teams and data access is recorded in an audit log. Each tool is built with its own Postgres schema and role, with queries running via an authenticated SQL gateway as that role.

We think Prized sits between products like Lovable and Retool. Lovable makes it easy to generate and host software, but it isn’t designed around distribution with permissions. Retool generally assumes that a technical builder is creating an app for an end user.

Prized treats internal tools as shared objects. Anyone in the workspace can see what others have built, fork, and connect different data. For example, one customer’s marketing lead built a promotional analytics tool. A data scientist at the same company then forked it and added confidence intervals with the existing tool as a starting base. This way workspaces become libraries of tools that people can reuse.

We’re live and self-serve. Our free tier includes 2 tool builds/month and our Teams tier is $100/month. The Enterprise tier is custom and supports personalized features like on-prem deployment.

We're still working out the right boundary between control and freedom. If you've built internal tools before we'd appreciate your feedback!

Comments

Comment by wseadowntown 4 days ago

I like the security-first posture. What's your view on how foundation models will or won't evolve into this space? Like will CC steamroll this in 2 years when it can natively build connectors and run them from within the desktop app? Not saying you won't have an ongoing edge, I just want to understand the thesis better so I can learn. Cool product!

Comment by hudsongr 4 days ago

Thanks! We believe the building aspect is already commoditized and we're experimenting with leaning into this with BYOA (bring-your-own-agent). Our bet is that the labs will stay focused on the model and won't do the more unglamorous governed deployment layer needs (credential scoping, per-tool db, audit, etc), but time will tell.

Comment by rimeice 4 days ago

Don’t know if anyone else is feeling this, but I’m getting a sense they’re gona have to go in to this layer. Not enough money to warrant the investment in the model only approach…

Comment by zormino 3 days ago

Particularly if LLMs plateau and intelligence becomes essentially commoditized. If they can't compete on models alone they'll start moving more and more up the stack.

Comment by dchuk 4 days ago

I’m playing with this exact concept but going down the path of a Claude code plugin that is optimized for configuring lowdefy apps because of that framework’s unique approach where you don’t write code you generate yaml configs that then drive the rendering of an app. So far it’s working pretty well, still pressure testing it. The audit control and permissions management you have is great, especially around the connectors. Nice job!

Comment by hudsongr 4 days ago

Thanks! We've actually have been experimenting with the config generation approach as well. One tradeoff we noticed was it turns the system into more of an app builder where the ceiling of what you can do is lower compared to a coding agent. But it's much faster than codegen. How dynamic have you found Lowdefy compared to codegen?

Comment by AnonHP 4 days ago

> Any call the agent makes to an external connector is reviewed by an LLM judge to prevent dangerous operations.

A few basic questions: how reliable is this judge since it’s based on an LLM? What additional measures can an admin or someone with more technical knowledge take to tighten this further if needed? Are the rules in or used by this judge visible to the users or an admin?

Comment by tcdent 4 days ago

It's not reliable, and presenting it as-such shows a lack of understanding of basic security posture.

Don't take my word for it, watch the news in <6 months and we'll certainly see it get bypassed.

Comment by dewey 4 days ago

Maybe that's a hot take, but it's probably still safer than how things are often done in reality when employees want to "get something done" and sign up for random services, download their own software or other variants of shadow IT.

Comment by tcdent 3 days ago

Completely agree.

This posture is applicable while we ease out of what I'm calling the "golden era" (loose restrictions on access with limited consequence) but one of the reasons I am pedantic and bearish about solutions like this is that they too have a short lifespan that we can already estimate the expiration date of.

Comment by ChrisMarshallNY 3 days ago

Aren't a large number of massive data leaks caused by some Marketing person setting up an AWS instance, and dumping the corporate DB into it?

Comment by dewey 3 days ago

I wouldn't say that's a "Marketing person" failure mode, there's enough open buckets, public mongodb or Elastic instances with wrong credentials.

Comment by ChrisMarshallNY 3 days ago

Fair point.

Comment by felixding 3 days ago

It's not safe or reliable, no matter how hard you try. Fundamentally, agents will try whatever they can to achieve the goal, as we've seen in many recent security incidents.

If determinism is important, as it is in many business use cases, use agents to build the tool, but don't let them run it.

I built a similar product around this exact idea: AI builds the workflow. Fixed rules run it.

Comment by hudsongr 3 days ago

We assume LLMs can always be fooled, so the judge is just one step of the verification process. The deterministic layers, per-tool db roles, read-only-by-default integrations, query timeouts, are the ones an admin can configure. The judge's rules aren't exposed to admins today. I like your idea of letting admins add their own rules or knowledge to the judge, we should add that.

Comment by Arshad-Talpur 3 days ago

One of the core issue here is LLMs are probabilistic not deterministic, even with set of rules or context , llms are inherently built to make best guess, so my question is how you are ensuring that LLMs are following rules?

Comment by marinoseliades 3 days ago

You're right, and we don't. The judge is best-effort screening not enforcement. Enforcement is deterministic with per-tool Postgres roles, proxy-injected creds, host allow-list, human approval on destructive writes. Those hold whether the judge is right or wrong.

Comment by iamniels 4 days ago

Congratulations with your launch! I suspect the product market fit for tools like these will be huge. Especially for SMBs with just 10s of employees in the office, lacking the budget for SAP consultants. However I think the agent creating an app is adding unnecessary complexity. Users want answers or insight in data, why build an app for that if the agent can provide it directly?

Comment by joelthelion 4 days ago

You can reuse the app indefinitely, for cheap, and predictably. I would say these are huge advantages.

Comment by hudsongr 4 days ago

Thanks! Agree that for one-off questions, chat is the right interface. We’ve found apps win when the workflow is repeated across a team and they almost become “templates” that seed ideas for other coworkers.

Comment by nodesocket 4 days ago

Congratulations on the launch. I’m very bullish on this concept of allowing SMB’s and startups to roll their own software and internal tools. Many many moons ago I built an entire inventory tracking and CRM for a very small company in PHP (circa 2006). It ran the entire company with orders, customers, devices, and revenue. It was a monstrous app, and took forever. Today, could be build in a matter of days and much better.

Along the same lines. I got so fed up with Quickbooks self employed (god awful software and company) for my own business I wrote a replacement using Cursor. It tracks expenses (categorizing each in proper IRS categories) and estimating quarterly taxes. Uses Plaid to pull in business bank accounts and credit cards and has income and expense reports. It’s 100x better than Quickbooks.

Comment by marinoseliades 3 days ago

Thanks! Custom beats generic if building it takes days instead of months.

Comment by Echo4309 4 days ago

Awesome launch and good pricing. We built an equivalent version in our org that takes HTML/JSX files and stores/serves them like an S3 bucket would. Works fabulous and we considered turning it into a SaaS product, met a real need in our org. Wishing you guys the very best of luck!!

Comment by IpV8 4 days ago

I am currently building something similar for my company. We are enabling people to host apps with python backends as well, so that people can do more complex tweaking of data for scientific workflows. The out-of-box solutions keep users constrained to an environment that doesn't allow things like importing weird packages on the back and front end for custom apps. The security implications are challenging, but I can mostly just keep the apps hosted in a highly constrained environment.

Separately I haven't yet seen a great security governance model for LLM integrations. At an enterprise level I'd like to govern rules such that, for example, if someone gives an LLM access PII information or proprietary data, then it shouldn't have access to a slack integration or the internet. Controlling this at the employee or team level doesn't make sense, as an employee may have reason to make separate use of both. From a security perspective I want them to be able use LLMs with different types of acccess, but not necessarily the same agent at the same time. Furthermore, they could ideally chain together agents with different permissions in specific orders. For example they could have a workflow where an agent can reach out to the internet, and then have a separate one that can read/write to slack, and then have a third that can interact with PII data. If they tried to wire together agents in the opposite order, it should get denied.

Comment by marinoseliades 3 days ago

We ended up splitting permissions per-tool for data and per-person for connections, for basically this reason. Every tool gets its own db role that can only see its own schema, and integration credentials live in a broker the tool's code never touches. So one tool can't read another tool's data, and nothing the model writes can leak a credential.

The ordering thing is harder and currently we don't do it. Per-tool permissions at best ban a combination, they can't say "internet then Slack is ok, the reverse isn't." DeepMind's CaMeL paper (arxiv.org/abs/2503.18813) is the best take I know on this, it tracks data flow between tool calls and enforces policy there. Still an open problem as far as I can tell.

Comment by marinoseliades 4 days ago

Thanks! That's the best kind of validation. If you ever take Prized for a spin, would genuinely love to hear how it compares to what you built, especially on the data access side.

Comment by aliclark 4 days ago

Me and my co-founder are building something similar but it started out as a tool to help build Lovable style prototypes for your existing product. Think Product Managers building an idea without having to figure out how to install git, VS Code, 17 other tools and then try share a URL.

Initial talks with customers resonated more with the internal tooling idea though. It's flexible enough to support both.

We're earlier along than this product though so there's a fair few wrinkles in our product.

https://flexenv.com/

Comment by renezander030 3 days ago

On the egress judge: an LLM reviewing connector calls inline is a probabilistic control sitting where a capability grant belongs. The operations that actually cause damage are largely enumerable: writes, deletes, and bulk reads of personal data. Those should be absent from the scoped token rather than present and then argued about by a judge with a nonzero false-negative rate under adversarial input, with the judge kept for the ambiguous long tail where enumeration genuinely fails. This is the WAF story again, request inspection in front of an app was a useful compensating control, and every team that mistook it for the boundary relearned that the boundary was the parameterized query and the database grant.

Comment by warthog 4 days ago

I did build a very similar product putting emphasis on integrations and security. It was tough to churn anyone out from Replit, Lovable and then CC and Codex.

When I talked to Lovable team, they told me Lovable is used primarily for prototyping, not internal tools and they are having a rough time trying to change the use case.

Hope my experience helps!

Comment by marinoseliades 3 days ago

Thanks for the feedback. From what we've gathered so far, non-eng are using CC or Codex to build one-off internal tools, but they're deploying them on their personal Vercel accounts and sending the url over Slack. That's the workflow we want to replace with a reasonable level of security and auditability. We're still trying to gauge interest in using the CLI to build and having everything deploy on Prized, we're working out whether that's valuable.

Comment by jedberg 4 days ago

How does this differ from retool?

Comment by themanmaran 4 days ago

TBH I was a retool power user for years, but I feel like they completely missed the vibe code train.

It's so much more annoying to go drag boxes around and type out little JS snippits in every chart as opposed to just telling claude to make the chart.

Comment by jedberg 4 days ago

Have you looked at their latest product? You describe your app in a text box and the AI builds it, with all the enterprise goodies like RBAC baked in.

Comment by marinoseliades 3 days ago

The difference is Retool still basically trusts whoever (or whatever) is building the app, we don't. The agent writing the code never sees a credential and each tool gets its own Postgres role. Even if it generates something dumb it physically can't touch data outside its scope.

Comment by big_hacker 4 days ago

It's funded by YC!

Comment by jedberg 4 days ago

So is retool. :)

Comment by rgbrgb 4 days ago

Very cool. What model is used for the judge? We use setoku for Claude-built internal tooling but the design doesn’t allow writes so that there’s no inference cost on the server. Having the judge check for danger is a neat design.

Comment by marinoseliades 4 days ago

Today it's GPT-5.6 Luna sitting inline on the egress broker, with Terra re-judging anything Luna flags as ambiguous. We are still experimenting though.

Comment by karam_hn 3 days ago

As someone who isn't an engineer but shipped a real product this year by leaning on AI tooling, the thing that surprised me most wasn't building the features — it was everything around them: payments, auth, not breaking prod, knowing when the output was subtly wrong. The "build" got cheap; the judgment didn't. Curious how you handle that last part for non-technical users.

Comment by Normal_gaussian 3 days ago

The basic build was never that expensive; the complications are always where the time goes.

Comment by blyry 4 days ago

nice this is awesome, what i've been building internally for a while now. lovable-esque internal platform with an agentic chat building experience, preview window, our internal auth, internal standards, sandboxed agent devcontainers on k8s. Even down to the shared experience -- these are shared tools for everbody, team ownership structure, not individual. easily forkable, auto-genned product docs etc etc. Hope you can monetize it! Look into private endpoint connectivity for big corps that need the internal access but don't want to self host.

Comment by marinoseliades 3 days ago

Thanks! The private endpoint tip is very much on point. We've already hit two prospects where the blocker was that they don't want full self-host but their security groups won't take generic cloud egress. It's on the list.

Comment by felixding 3 days ago

Please don't let users "try" your product only to immediately hit a signup/signin wall. It instantly undermines whatever trust the demo may have established.

Comment by marinoseliades 3 days ago

More on this, we're still experimenting between a hard vs. soft paywall. Regardless, if you sign up it should be easy enough to play around with it for a while on the free tier. We promise that we don't send spam/marketing emails.

Comment by felixding 4 hours ago

It’s not about what you offer on the free tier. It’s about whether what you say matches what you do.

If your page asks people to try the product, let them try it. Asking them to enter something and click “Try,” only to immediately hit a signup wall, is frustrating and makes them feel tricked.

Comment by scottydelta 3 days ago

I know someone who could use this but a price point/plan between $0 and $100 would be a good way to help people get started.

Comment by marinoseliades 3 days ago

Heard, we're actively working on this and should have updated pricing plans soon.

Comment by Suppafly 1 day ago

>Let non-engineer staff build secure internal tools

That seems like a really bad idea.

Comment by christoff12 4 days ago

I like this a lot. Do you handle scenarios where the data and context live in a data warehouse (Snowflake, etc)?

There are a lot of operational needs that exist downstream from dbt models that combine data from the various source systems.

Comment by hudsongr 4 days ago

Yes, we support Snowflake and ClickHouse today. And we're happy to integrate more, just say the word :)

Comment by bitlad 4 days ago

You can do this in lovable too now. But i dont think it gets used a lot. I think it is not very unique, i have seen bunch of providers, some are selling ai workers, some are selling internal tools in different shape and form.

I think the moat for internal tools is not where you think it is. Also, i see self host option, it is a nightmare too, no one will do it just for internal tools now. It was valid before AI era.

I am also seeing non engineering staff actually building tools with claudecode, paas like railway and stuff. People have become more efficient with engineer tools.

Comment by marinoseliades 3 days ago

Fair, the generation part is commoditized. What we sell is the trust layer around company data. No credentials in the builder's hands, scoped access per tool, forkability etc. CC + Railway works until someone pastes a prod key into an env var.

Comment by bitlad 3 days ago

This is not the most effective way to build trust layer. But anyways.

Comment by fabijanbajo 3 days ago

the credentials-out-of-the-sandbox part is the easy half. the quieter failure is a tool that holds no keys and still renders rows this particular employee was never meant to see.. sign-in proves who someone is, it doesn't scope what the query returns.

does the scoped token narrow to the requesting user, or is it one service identity per connector with the row filtering left to whatever the agent generated?

Comment by inflam52 4 days ago

Seems like Superblocks

Comment by marinoseliades 3 days ago

Closest comp. I'd argue the "AI builds the app part" is fast becoming table stakes. The unsolved problem is what happens after the app exists and that's what we actually built. Anyone in the org can fork a coworker's tool and rebind it to their own data scope. Enforcement also happens at a different layer. From what we've gathered, Superblocks controls who can run an app and bakes guardrails in when the app is generated, but every app still queries through a shared database credential. On our side, each tool gets its own database role that can only see its own data, and credentials live in a proxy the code never touches. So even when the AI writes bad code, it can't reach anything it shouldn't.

Comment by Bnjoroge 4 days ago

Demo looks cool, congrats on the ship! Probably missed but is it possible to externally share/view the built artifact outside of prized?

Comment by hudsongr 4 days ago

Thank you! Yes, you can share it externally if it isn't connected to external data sources.