alt.hn

12/8/2025 at 5:10:14 PM

Launch HN: Nia (YC S25) – Give better context to coding agents

 https://www.trynia.ai/

by jellyotsiro

12/8/2025 at 6:00:16 PM

Cursor promises to do this[0] in the product, so, especially on HN, it'd be best to start with "why this is better than Cursor".

> favorite doc sites so I do not have to paste URLs into Cursor

This is especially confusing, because cursor has a feature for docs you want to scrape regularly.

0 - https://cursor.com/docs/context/codebase-indexing

by mritchie712

12/8/2025 at 6:02:12 PM

The goal here is not to replace Cursor’s own local codebase indexing. Cursor already does that part well. What Nia focuses on is external context. It lets agents pull in accurate information from remote sources like docs, packages, APIs, and broader knowledge bases

by jellyotsiro

12/8/2025 at 6:37:33 PM

That’s what GP is saying. This is the Docs feature of Cursor. It covers external docs/arbitrary web content.

`@Docs` — will show a bunch of pre-indexed Docs, and you can add whatever you want and it’ll show up in the list. You can see the state of Docs indexing in Cursor Settings.

The UX leaves a bit to be desired, but that’s a problem Cursor seems to have in general.

by jondwillis

12/8/2025 at 6:41:02 PM

yeah ux is pretty bad and overall functionality. it still relies on a static retrieval layer and limited index scope.

+ as I mentioned above there are many more use cases than just coding.Think docs, APIs, research, knowledge bases, even personal or enterprise data sources the agent needs to explore and validate dynamically.

by jellyotsiro

12/8/2025 at 8:25:54 PM

As an AI user (claude code, rovo, github copilot) I have come across this. In code it didnt build something right where it needed to use up to date docs. Luckily those people have now made an MCP but I had to wait. For a different project I may be SOL. Suprised this isnt solved, well done for taking it on.

From a business point of view I am not sure how you get traction without being 10x better than what Cursor can produce tomorrow. If you are successful the coding agents will copy your idea and then people being lazy and using what works have no inventive to switch.

I am not trying to discourage. More like encourage you to figure out how you get that elusive moat that all startups seek.

As a user I am excited to try it soon. Got something in mind that this should make easier.

by nrhrjrjrjtntbt

12/8/2025 at 8:37:08 PM

thanks! will be waiting for ur feedback

by jellyotsiro

12/8/2025 at 7:16:12 PM

This is different because of the background refresh, the identifier extraction and the graph. I know because I use cursor and am building the exact same thing oddly enough.

by bn-l

12/8/2025 at 5:47:11 PM

This looks neat, we certainly need more ideas and solutions on this space, I work with large codebases daily and the limits on agentic contexts are constantly evident. I've some questions related to how I would consume a tool like this one:

How does this fare with codebases that change very frequently? I presume background agents re-indexing changes must become a bottleneck at some point for large or very active teams.

If I'm working on a large set of changes modifying lots of files, moving definitions around, etc., meaning I've deviated locally quite a bit from the most up to date index, will Nia be able to reconcile what I'm trying to do locally vs the index, despite my local changes looking quite different from the upstream?

by 6thbit

12/8/2025 at 5:54:08 PM

great question!

For large and active codebases, we avoid full reindexing. Nia tracks diffs and file level changes, so background workers only reindex what actually changed. We are also building “inline agents” that watch pull requests or recent commits and proactively update the index ahead of your agent queries.

Local vs upstream divergence is a real scenario. Today Nia prioritizes providing external context to your coding agents: packages, provider docs, SDK versions, internal wikis, etc. We can still reconcile with your local code if you point the agent at your local workspace (cursor and claude code already provide that path). We look at file paths, symbol names and usage references to map local edits to known context. In cases where the delta is large, we surface both the local version and the latest indexed version so the agent understands what changed.

by jellyotsiro

12/9/2025 at 1:33:51 PM

Your FAQ says you don’t store code. But this answer sounds like you do? Even if you’re storing as an embedding that’s still storage. Which is it?

by adam_patarino

12/9/2025 at 6:28:07 PM

We don’t store your code or any proprietary local content on our servers. When we say “external context” we mean public or user-approved remote sources like docs, packages or APIs. Those are indexed on our side. Your private project code stays local

by jellyotsiro

12/8/2025 at 10:52:49 PM

This resonates. I'm building a React Native app and the biggest friction with AI coding tools is re-explaining context every time.

How does Nia handle project-specific patterns? Like if I always use a certain folder structure or naming convention, does it learn that?

by alex-ross

12/8/2025 at 10:58:40 PM

Nia is focused on external context rather than learning the patterns inside your own codebase. Cursor and IDE-native tools are better for local project structure today. Where Nia helps is when the agent needs ground truth from outside your repo. For example, you can index React Native docs, libraries you depend on, API references or Stack for your backend and let the agent search and validate against those sources directly instead of losing context between prompts.

by jellyotsiro

12/8/2025 at 6:43:53 PM

I've no idea what their architecture/implementation looks like, but I've built a similar tool for my own use and the improvements are dramatic to say the least.

Mine's a simple BM25 index for code keyword search (I use it alongside serena-mcp) and for some use cases the speeds and token efficiency are insane.

https://gitlab.com/rhobimd-oss/shebe#comparison-shebe-vs-alt...

by marwamc

12/8/2025 at 7:32:50 PM

It's amazing that you can do this on your own but the company in question cannot although they have raised $6M. If your README were their entire pitch and website and, it would be 100000x more convincing.

by ModernMech

12/8/2025 at 9:12:10 PM

YC these days is just a fast track for rich kids to raise millions with nothing but a landing page.

Vaporware.

by bluerooibos

12/8/2025 at 6:59:35 PM

(I detached this subthread from https://news.ycombinator.com/item?id=46195560 because your comment is fine and I don't want it to be penalized.)

by dang

12/8/2025 at 8:19:52 PM

Thanks @dang

by marwamc

12/9/2025 at 6:05:36 AM

perhaps you should thing of launching a company to improve on hour tool, yc or not.

by Alive-in-2025

12/10/2025 at 2:22:49 PM

Looks cool, it is always a good take to make agent to retrieve information well- writing docs every time can always go bad.

by mengmota

12/11/2025 at 3:28:54 AM

thanks! yes writing docs is pain the ass haha

by jellyotsiro

12/9/2025 at 12:35:32 PM

> The calling agent then decides how to use those snippets in its own prompt.

To be reductionist, it seems the claimed product value is "better RAG for code."

The difficulties with RAG are at least:

1. Chunking: how large and how is the beginning/end of a chunk determined

2. Given the above quote, how much or many RAG results are put into the context? It seems that the API caller makes this decision, but how?

I'm curious about your approach and how you evaluated it.

by djoldman

12/9/2025 at 6:29:01 PM

Not quite “better RAG for code”. The core idea is agentic discovery plus semantic search. Instead of static chunks pushed into context, the agent can dynamically traverse docs, follow links, grep for exact identifiers, and request only the relevant pieces on demand.

No manual chunking. We index with multiple strategies (hierarchical docs structure, symbol boundaries, semantic splitting) so the agent can jump into the right part without guessing chunk edges.

Context is selective. The agent retrieves minimal snippets and can fetch more iteratively as it reasons, rather than preloading large chunks. We benchmark this using exact match evaluations on real agent tasks: correctness, reduced hallucination, and fewer round trips.

by jellyotsiro

12/8/2025 at 7:17:34 PM

Congrats on the launch. The problem is definitely there. I wonder how are you planning to differentiate yourself from Cursor and the like. You mention you are complementary, but Cursor provide similar features to add external doc context for instance to a prompt. I understand you do better in your benchmark, but with the amount of funding they may be able to replicate and improve over it (unless you have a secret thing).

by canopi

12/8/2025 at 7:21:08 PM

as I mentioned above there are many more use cases than just coding (APIs, research, knowledge bases, even personal or enterprise data sources the agent needs to explore and validate dynamically)

I started out with coding agents specifically because it came from personal pain of how horrible they are with providing up to date context.

by jellyotsiro

12/8/2025 at 8:21:08 PM

This looks interesting and worthwhile. I did a double-take when I read "when (a few months ago) I was still in high school in Kazakhstan"

by chrisweekly

12/9/2025 at 2:24:30 AM

Configure MCP Server One command to set up Nia MCP Server for your coding agent.

Select your coding agentCursor Installation method Local Remote Runs locally on your machine. More stable. Requires Python & pipx.

Create API Key test Create Organization required to create API keys

i can not create api key? the create button is grey and can not be pressed.

by qcqcqc

12/9/2025 at 3:51:03 AM

hey, what error does it throw?

by jellyotsiro

12/8/2025 at 5:37:54 PM

Having this RAG layer was always another thing to try for me. I haven't coded it myself, and super interested if this gives a real boost while working with Claude. Curious from anyone who have already tried the service, what's your feedback? Did you feel you're getting real improvements?

by RomanPushkin

12/8/2025 at 5:41:07 PM

Wouldn’t call it just RAG though. Agentic discovery and semantic search are the way to go right now, so Nia combines both approaches. For example, you can dynamically search through a documentation tree or grep for specific things.

by jellyotsiro

12/8/2025 at 5:44:23 PM

We call it agentic RAG. The retriever is an agent. It’s still RAG

by zwaps

12/8/2025 at 5:46:45 PM

Which would be much better than the techniques used in 2023. As context windows increase, combining them becomes even easier.

There are a lot of ways of how you can interpret agentic rag, pure rag, etc

by jellyotsiro

12/9/2025 at 7:24:29 PM

Can you explain why I would pay almost the full price of Cursor, ChatGPT, or Claude again - just for your context layer, when these companies are already working on context?

I don't see you justify this with an explanation of the ROI anywhere.

by bluerooibos

12/12/2025 at 4:08:46 PM

Just like with tool calling, I find models that are post-trained to work with a context system use it more reliably. I don't think bolting on a context system is ever going to work as well.

by adam_patarino

12/9/2025 at 8:35:29 PM

The economics are simple. When an agent guesses, it produces wrong code, failed runs, and wasted time. External context is the biggest source of those mistakes, because IDEs only index what’s in your repo. We are a complement to Cursor, ChatGPT, and Claude, not a replacement.

by jellyotsiro

12/8/2025 at 5:45:40 PM

The context problem with coding agents is real. We've been coordinating multiple agents on builds - they often re-scan the same files or miss cross-file dependencies. Interested in how Nia handles this - knowledge graph or smarter caching?

by govping

12/8/2025 at 5:49:39 PM

hey! knowledge graphs are also used at runtime but paired with other techniques, since graphs are only useful for relationship queries.

by jellyotsiro

12/8/2025 at 7:17:48 PM

Is the RAG database on your servers or is it local? If not local is there a local option?

by bn-l

12/8/2025 at 7:20:06 PM

hey! i use multiple DBs but the primary ones are turbopuffer and chroma for package search. they are really great

re local, I do local for certain companies!

by jellyotsiro

12/9/2025 at 8:24:18 AM

Nice. Huge congrats on the launch!!!

by bn-l

12/9/2025 at 7:18:59 PM

Your landing page tells me a whole lot of nothing.

How does this work? How does it differ from other solutions? Why do I need this? What does the implementation look like if I added this to my codebase?

by bluerooibos

12/9/2025 at 10:08:04 AM

So many coding tools what makes you different.

by dhruv3006

12/9/2025 at 6:29:48 PM

We’re not a coding tool. We’re the context layer for agents.

by jellyotsiro

12/9/2025 at 9:53:56 AM

Very happy to see this since I am building in this domain. We need external and internal context though. I am aiming for always available context for current and related projects, reference projects, documentation, library usage, commands available (npm, python,...), tasks, past prompts, etc. all in one product. My product, nocodo (1), is built by coding agents, Claude Code (Sonnet only) and opencode (Grok Code Fast 1 and GLM 4.6).

I just made a video (2) on how I prompt with Claude Code, ask for research from related projects, build context with multiple documents, then converge into a task document, shared that with another coding agent, opencode (with Grok or GLM) and then review with Claude Code.

nocodo is itself a challenge for me: I do not write or review code line by line. I spend most of the time in this higher level context gathering, planning etc. All these techniques will be integrated and available inside nocodo. I do not use MCPs, and nocodo does not have MCPs.

I do not think plugging into existing coding agents work, not how I am building. I think building full-stack is the way, from prompt to deployed software. Consumers will step away from anything other than planning. The coding agent will be more a planning tool. Everything else will slowly vanish.

Cheers to more folks building here!

1. https://github.com/brainless/nocodo 2. https://youtu.be/Hw4IIAvRTlY

by brainless

12/8/2025 at 6:32:43 PM

Benchmarks?

by orliesaurus

12/8/2025 at 6:37:36 PM

Arlan had this in his text but I cut it for brevity - sorry about that! Here's the bit:

In our internal benchmark on bleeding edge SDK and library features, Nia produced the lowest hallucination rate among the context providers and search tools we tested (context7, exa code, etc), and I wrote up the setup and results in a separate blog post: https://www.nozomio.com/blog/nia-oracle-benchmark

by dang

12/9/2025 at 10:14:53 AM

Is this similar to the indexing done by Jetbrains IDEs?

by krisgenre

12/9/2025 at 6:30:27 PM

JetBrains is great at indexing your local codebase and understands it deeply. We don’t try to replace that. Nia focuses on external context: docs, packages, APIs and other remote sources that agents need but your IDE can’t index.

by jellyotsiro

12/8/2025 at 9:59:40 PM

SOTA on internal benchmark?

by jacobgorm

12/8/2025 at 10:05:51 PM

going to open source it soon :)

by jellyotsiro

12/8/2025 at 7:44:17 PM

Congrats. From my experience, Augment (https://augmentcode.com) is best in class for AI code context. How does this compare?

by kenforthewin

12/8/2025 at 7:46:12 PM

augment is a coding agent. nia is an external context engine for coding agents that improves their code output quality

by jellyotsiro

12/8/2025 at 9:09:06 PM

Sure, but Augment’s main value add is their context engine, and imo they do it really well. If all they had to do was launch an MCP for their context engine product to compete, I think the comparison is still worth exploring.

by kenforthewin

12/8/2025 at 8:35:00 PM

How does it compare to Serena MCP? :)

https://github.com/oraios/serena

by mike1505

12/8/2025 at 8:36:09 PM

Serena is great for semantic code editing and symbol-level retrieval on your own codebase. It gives the agent IDE-like capabilities inside the repo. Nia focuses on a different layer. We target external context: remote code, docs, packages, APIs, research, enterprise knowledge, etc

w nia the agent can dynamically search, traverse, and validate information outside the local project so it never hallucinates against out-of-date or incomplete sources.

by jellyotsiro

12/8/2025 at 5:42:38 PM

Absolutely insane that we celebrated coding agents getting rid of RAG, only with the next innovation being RAG

by zwaps

12/8/2025 at 5:44:30 PM

Not exactly just RAG. The shift is agentic discovery paired with semantic search.

Also, most of the coding agents still combine RAG and agentic search. See cursor blog about how semantic search helps them understand and navigate massive codebases: https://cursor.com/blog/semsearch

by jellyotsiro

12/8/2025 at 6:09:57 PM

This is happening over and over and over. The example of prompt engineering is just a form of protocol. Context engineering is just about cache management. People think LLMs will replace programming languages and runtimes entirely, but so far it seems they have been used mostly to write programs in programming languages, and I've found they're very bad interpreters and compilers. So far, I can't really pick out what exactly LLMs are replacing except the need to press the individual keys on the keyboard, so I still struggle to see them as more than super fancy autocomplete. When the hype is peeled away, we're still left with all the same engineering problems but now we have added "Sometimes the tool hallucinates and gaslights you".

by ModernMech

12/8/2025 at 5:43:41 PM

The pendulum swings back.

by choilive

12/9/2025 at 1:42:37 PM

[dead]

by Jakedismo

12/8/2025 at 7:00:38 PM

[under-the-rug stub]

[see https://news.ycombinator.com/item?id=45988611 for explanation]

by dang

12/8/2025 at 7:09:16 PM

congrats on the launch Arlan! Nia is a lifesaver when we're coding :)

by ayushrodrigues

12/8/2025 at 7:05:15 PM

Congrats on the launch! Definitely a problem I’ve run into myself.

by rushingcreek

12/8/2025 at 7:44:40 PM

Congrats on the launch, Nia looks great.

by agentastic

12/8/2025 at 6:00:04 PM

super smart. congrats on the launch!

by johnsillings

12/8/2025 at 5:59:37 PM

I love Nia, keep it up Arlan

by himike

12/8/2025 at 6:02:30 PM

thank you haha!

by jellyotsiro

12/8/2025 at 6:03:21 PM

Love it.

by ramzirafih

12/8/2025 at 6:55:28 PM

Amazing product. As an individual heavy user I must say it does what it is supposed to do and it does it well.

by phegler

12/8/2025 at 6:58:47 PM

glad to hear that!

by jellyotsiro

12/8/2025 at 6:05:00 PM

[flagged]

by ModernMech

12/8/2025 at 6:59:06 PM

Please make your substantive points without being snarky or aggressive. You have a good point in there at the end, but the site guidelines ask you not to comment like this, and that's especially important in Show or Launch threads.

https://news.ycombinator.com/newsguidelines.html

by dang

12/8/2025 at 7:24:52 PM

Ok here is my less snarky and less aggressive take on the site and product: it doesn't leave me with confidence, it makes me feel uneasy about them as a company, it makes me not trust them, and it makes me feel like I'm being lied to. To fix this, provide proof. Otherwise, stop making the claims. Unfortunately with the frequency these kinds of products come out of YC, it seems like maybe YC coaches these companies as if this is an effective way to communicate.

by ModernMech

12/8/2025 at 9:40:55 PM

The only coaching going on with Launch HNs is from me and tomhow, and I can tell you that we're constantly urging people to tone down grand claims, provide concrete examples, accessible demos, and so on—partly because it makes the posts more interesting, and partly to reduce surface area for the snarky and rigid objections that internet forums optimize for.

We don't do a perfect job of this, because (1) Launch HN coaching is on top of our main jobs running HN and we only have so many hours; and (2) startup founders' priority is working on their startup (as it should be!). They only have so many cycles for reworking everything to suit HN's preferences, which are idiosyncratic and at times curmudgeonly or cynical. Curmudgeons and cynics can't be convinced in the first place so it's not a good idea for a founder to put too much time into indulging them.

Some of what you're saying here boils down to that their home page shouldn't have any marketing tropes at all (e.g. testimonials, companies-using-us, etc.). I don't like those tropes either, but this is an example of what I mean by an idiosyncratic preference. Companies do that kind of thing because, obviously, it works. That's how the world is. The only thing that you accomplish by angrily blaming a startup founder for doing standard marketing is to make the discussion dyspeptic and offtopic. And yes, I do use the word "dyspeptic" too much :)

by dang

12/8/2025 at 10:44:53 PM

I mean, you can dismiss me by calling me a cynical curmudgeon (which is not inaccurate), but from my perspective their website doesn't try to convince -- it tries to bamboozle. I don't think it's idiosyncratic at all to expect that claims made should be proven and supported, and that companies should present themselves with integrity and be genuine in their representations.

> Curmudgeons and cynics can't be convinced in the first place so it's not a good idea for a founder to put too much time into indulging them.

I'd say we're just not convinced by marketing lingo and puffery. I was convinced by the simple README containing code and transparent evidence that a fellow HNer put up in their personal capacity, so maybe you can direct the Nia team to that as an example of how to properly convince curmudgeons and cynics.

by ModernMech

12/8/2025 at 11:35:06 PM

Kudos to you for "which is not inaccurate" and the subtle shift to "we" - that made me smile.

Personally my tastes are much the same as yours, but we're asking for too much if we want startups to stop doing normal marketing.

by dang

12/8/2025 at 7:41:37 PM

i am interested in knowing what would be the correct way to do it according to your checklist. For example you said testimonial from twitter can be bots, which testimonial according to you would give you confidence that product is genuine?

by pdyc

12/8/2025 at 9:12:00 PM

Okay let's dissect it. First, I will say since this is a for-profit corporate website and they are trying to get something from me, I approach it with a fully skeptical, 0 benefit of doubt perspective.

What's the gif supposed to tell me? It's supposed to demo the product and give me a feel for its capabilities. But it just flits around and goes so fast, offers zero explanation for anything, it just leaves me disoriented. So at minimum, this needs captions and it needs to go about 2x slower. But really, this one GIF should not be the most substantive element on the first page relating to the actual product and what it does. Trust lowered.

Moving on to the "company carousel", which is trying to say "these other companies trust us so you should too". They're trying to ride on the reputations of Stanford, Cornell, Columbia, UPenn, Google, etc. as a sort of pseudo-endorsement, because they cannot post real endorsements from these institutions, because they do not exist (doesn't YC have legal counsel to tell them this is illegal?). How are engineers using Nia at Stanford? We don't know, Nia will not say, likely because no one at Stanford is using it in any real capacity that is impressive enough to put on the front page of the website. If they were, then why wouldn't Nia tell us about that rather than just flashing the Stanford logo? So the logo suffices, and I guess the more logos the better. Trust lowered.

Next the investor list: who is this for and what does it communicate? It appears to be a list of Chiefs, VPs, Co-Founders, and various funds who are deemed to be "world class", which is just another parade of logos but for a different audience, likely other investors who know these people. Maybe this speaks to some people in terms of the project having a solid financial backing but that's a smokescreen to distract you from the fact there's no actual business plan here aside from running on the VC treadmill and hoping to get acquired by one of your customers and/or investors. Trust lowered.

Then we get to the Twitter parade, which is a third instance of "just trust us bro". And it includes such gems as "Can confirm, coding agents go hard" and "go try Nia, go into debt if you have to". Testimonials are for products I can't try myself, this seems like something that can be demoed, so why isn't it? Why did they opt to devote all this space to show a bunch of random people saying random uninteresting things about their product, rather than use the space to say more interesting things about their product? Because the testimonials are a distraction from the actual product. Trust lowered.

Again, I'm left asking: Why do I have to listen to and trust these other people if the technology is so good? Why am I halfway down the page reading this thing, and I've yet to hear any specifics about how this thing works or what it does for me. I was told other people are using it but not how, I was told other people invested in it but not how much, and I was told some companies are maybe using it but not in what capacity.

So in summary, this page is: "Look how shiny! You trust us. No really, you can trust us! Seriously, look at all these people, who say you can trust us, you seriously can! Now give us money."

So to answer your question:

> what would be the correct way to do it according to your checklist.

Don't do any of the things that were done, and instead lead with the product. Prove all claims made. If a claim can't be proven don't make it. Stand behind your technology rather than testimonials.

by ModernMech

12/9/2025 at 1:21:38 AM

Well put. But what you’ve identified here is pretty much the norm for ~90% startup/SAAS sites. They all just regurgitate this formula.

by replwoacause

12/9/2025 at 4:17:07 AM

Maybe worth considering whether this formula is still effective.

by akcho

12/8/2025 at 11:12:58 PM

@dang now that he has made his point, shouldn't the thread be made undead?

by fazkan

12/9/2025 at 1:22:52 AM

My snark shall never be resurrected, lest it thirst for brains.

by ModernMech