Do you still need an AI agent framework?
Often no. A framework helps you start — it gives you tool-calling, state, and orchestration out of the box — but as the model providers fold those primitives into their own SDKs, the framework's value shrinks. The durable advantage isn't the framework; it's your context: what data you retrieve, how you manage it, and what the system remembers. Many teams start on a framework and then go framework-light as their needs get specific.
What a framework actually gives you
Agent frameworks bundle the plumbing: tool/function calling, a loop, state passing, retries, and often pre-built patterns for multi-agent work. Early on that’s a real accelerant — you skip a lot of boilerplate and get to a working agent fast.
Why the value is shrinking
Two things erode it. First, the model providers keep absorbing those primitives into their own SDKs, so the framework is wrapping capabilities you increasingly get natively. Second, frameworks impose abstractions, and as your use case gets specific, those abstractions start fighting you — you spend time working around the framework instead of with it. That’s the “framework era is over” argument: not that frameworks are useless, but that they’re no longer where the hard, differentiating work lives.
Where the moat moved
The advantage moved to context. As frontier models converge and the plumbing commoditizes, what separates a good agent from a bad one is the quality of what you put in front of it — retrieval, context management, and memory — not which framework wired it up. That’s the part you own and the part worth investing in.
The practical read
Start with a framework if it gets you moving; there’s no prize for hand-rolling the loop. But don’t treat the framework as the product. Expect to go framework-light as your requirements sharpen, and put your real effort into context and evaluation, which is what actually determines whether the agent works.
From the conversation
This explainer is drawn from these episodes — each carries its full transcript.