
Why are there so many AI frameworks?
Why does it feel like every developer in the bay area announces ANOTHER MCP library to write client/servers for AI...the USB-C of AI they say, it feels to me like the nightmare of AI, because everything moves so fast and everyone is building more stuff (and barely maintaining it)
There are a lot of AI frameworks to begin with: LangChain. LlamaIndex. AutoGen. CrewAI. DSPy. OpenDevin. Semantic Kernel... Theyโre all trying to answer one basic question:
'The best way to reliably interact with an LLM so that you can bolt things like memory, tools, and MCP to your AI agent'
Everyone Wants to Be the `express.js` of Agents, the `node-fetch` of web-scraping or the `bun of agentic runtimes`
You walk around the bay area and you can see developers grinnin'... for protocol dominance.
Just like HTTP standardized how browsers talk to servers, every MCP framework is someoneโs moonshot to standardize how we feed, prompt, tool, etc in these models.
It took 30 years before we got the HTTP of today โ 30 years of computing standards and a clear abstraction (a document on a server).
We donโt have that clarity with AI and LLMs.
What we have now are probabilistic agents, infinite contexts, ephemeral memories, and LLMs that hallucinate confidently, making up stories.
So each platform has a different belief system, each one with a large group of fans backing them up:
LangChain says: break the world into tools and chains.
LlamaIndex says: structure the context from your data.
Toolhouse says: expose agentic calls cleanly, then get out of the way.
AutoGen says: prompt engineers should be prompt diplomats.
So...๐๐ต๐ ๐๐ผ ๐บ๐ฎ๐ป๐ ๐๐ฟ๐ฎ๐บ๐ฒ๐๐ผ๐ฟ๐ธ๐ ๐ฎ๐ป๐ฑ ๐ฃ๐น๐ฎ๐๐ณ๐ผ๐ฟ๐บ๐?
Weโre still early. Because everyone wants a piece of the pie.
Look how many MCP related products exist already: https://www.producthunt.com/search?q=MCP
No one trusts the abstractions they didnโt write themselves.
Every framework is a political stance, a philosophy on how to do the "NEW THING" right:
Where does context ends and reasoning begins?
How much to expose vs. hide to the final user?
Whether orchestration should be declarative or emergent.
Frameworks multiply because philosophies diverge.
Replies
Can't disagree.
Some of them make sense though.
Langgraph -> flexibility
letta -> Memory
CrewAI -> Simplicity
Depending on the USP might be useful.