The agentic AI space has grown so fast that keeping track of what exists is now its own project.
A publicly shared spreadsheet is making the rounds that attempts to map the entire agentic AI stack as of 2026 - from orchestration frameworks and memory layers to tool-use platforms and deployment infrastructure. The document categorizes vendors and open-source projects across multiple layers of what it takes to build AI agents (autonomous programs that can plan, use tools, and complete multi-step tasks without constant human input).
The timing makes sense. Over the past year, agentic AI has gone from a research concept to something companies are actively shipping. Salesforce has Agentforce, Microsoft has Copilot agents, and dozens of startups are building the picks-and-shovels layer underneath. But the space is fragmented enough that no single person can keep it all in their head anymore.
Maps like this are useful for teams evaluating their stack, but they come with a caveat: the agentic space is moving fast enough that any static snapshot starts aging the moment it's published. Tools merge, pivot, or disappear on a monthly basis. Still, as a starting point for understanding what categories of tooling exist and who the current players are, a structured overview beats scattered blog posts and vendor marketing pages.