Every founder tool shipped 'agents' in the last 12 months - here's what actually changed and what was a rebrand

Twelve months. Notion Agents, Notion Workers, Linear for Agents, Cursor multi-agent workspaces, Claude Code, ChatGPT Operator, Anthropic Vercept, Vercel v0 IDE - every tool in the founder stack now ships an 'agent' surface. The honest read: maybe a third of these are genuinely new capability, the rest are LLM features rebranded as agents to hit the marketing moment. Real new capability looks like Linear's Agent API (read/write issues programmatically), Notion Workers (agents can call external APIs), Cursor Composer (a coding model trained for autonomy), Operator and Vercept (browsers that click). Rebrand looks like 'AI assistant' that always existed now sitting under an Agents tab. If you're choosing tools right now, ask one question: does the agent feature change what's possible, or just what's labelled? If it's labelling, ignore it. If it changes what's possible - particularly for tasks you'd previously have to either do yourself or wire up Zapier for - it deserves an evening of testing.