Ask HN: Why Is Every Company Building Own Agent Framework? Isn't One Enough?
It feels like every major AI company — OpenAI, Anthropic (Claude), Google, etc. — is now rolling out its own agentic framework, orchestration library, or tool for managing LLM “agents.” We’ve seen LangChain and similar projects actively evolving in this space, but now each big player is launching something new almost monthly.
As a developer and researcher, I’m wondering: • Why are we reinventing the wheel so often instead of improving a common SDK or open framework? • Is there a fundamental technical reason (e.g., architectural, security, integration) why companies don’t just contribute to or extend existing libraries? • Or is it more about platform lock-in, developer ecosystems, or strategic positioning?
Would love to hear perspectives from people working in or close to these projects.
It's way too early in this technology maturity lifecycle for this to be a prime focus. Despite the naysyers, innovations in this space are happening on a weekly if not daily pace.
Matirity, convergence and eventual standardization will organically arise down the line as the space settles down, but that time is not now.
Why do we have php, .net, Django?
Why react, angular, whatever else is front end?
Why different clouds and similar services?
Leader, front runner, early adopter and not invented here mindset allows competition and community to rally around what will become the next big thing
That’s a fair point — ecosystems naturally evolve with competing frameworks, and sometimes that diversity helps the best ideas emerge. But in those examples (PHP vs. .NET vs. Django, React vs. Angular), we usually see clear differences in philosophy, performance, developer experience, or use case fit.
What I find puzzling in the current wave of AI agent libraries is that many seem to be doing almost exactly the same thing — wrapping LLMs, defining tools, planning steps — just with slightly different APIs or names. It’s hard to see a fundamental divergence in capability or vision, at least so far.
So my question is less about why variety exists — and more about whether this current fragmentation is actually productive, or just noise driven by branding and platform lock-in strategies.
Would love to be wrong here if others are seeing more substantial innovation beneath the surface.
Also, I’d love to see a shared protocol or framework emerge — something like the Model Context Protocol (MCP) or an open standard — that different tools and platforms can build on and improve together. That kind of foundation could push the whole ecosystem forward instead of scattering innovation across isolated silos.
roight but not every company is making their own language / framework
Yes I hear you. I think though that controlling the interface to your platform controls market share, integration, pricing and other business benefits so while everything looks the same at the tech level it’s the value of lockin that is the driver. I may be wrong
It was like the search engine era. Everyone threw money at everything and Google won. Looks like Cursor might be the next Google at this rate. But there's still other things to win at in the AI arena, so there's probably some FOMO.
I do believe there's room for segmentation on agents as well. Manus and ChatGPT Deep Research are fairly different and they likely access different sources. Grok seems to play mostly on social media and so on. I think companies need the kind of depth that Cursor has. GitHub tried to do a bit of everything and failed despite their resources.
I don't know about other companies or even other teams, but given the chance I'd definitely build my own stuffs. It's a LOT more interesting than using other people's stuffs.
It's just in my blood. I want to build things that others use, and build it really well with a perfect documentation and some hand-on videos.
A few options on the landscape is usually a good thing.
> Why are we reinventing the wheel so often instead of improving a common SDK or open framework?
Welcome to the world of software technology, can I interest you in one of the 87 JavaScript frameworks out there that all manipulate an HTML DOM in one way or another?
Setting the sarcasm aside, I think there are "AI" companies popping up which do these types of things - orchestration, agent libraries, etc. so it makes some sense for each AI company to have their own flavor of those available.
Edit to add: I put "AI" company in quotes because at a technical level these companies/apps aren't creating AI themselves, they are wrapping AI HTTP APIs.
all of them are bad and force software 1.0 principles on agents for no reason
https://xkcd.com/927/