We have a problem, and it isn’t a shortage of tools. There are now more than 650 tools, standards and services competing across 36 layers of the agentic stack, and the list changes every month. Frameworks get renamed, startups get acquired, repos get archived, and protocols change hands. Nobody can keep the full map in their head, so the hardest question has become “which one should we use, and is it still the right one next quarter?”

Why is choosing so hard

Overlap everywhere. Dozens of frameworks, vector databases, sandboxes, and gateways claim to solve the same problem, each with a different license, language, and maturity level.

Fast decay. A tool that was the default six months ago may now be in maintenance mode or absorbed by a larger vendor. Facts go stale faster than we can read them.

Hidden trade-offs. “Open source” can mean permissive, copyleft or source-available with commercial limits. The license decides whether we can ship.

Wrong question. Most teams ask, “What’s the best tool?” The better question is “which layer is this, and what does it have to survive?”