AI agents moved from demo to production faster than most teams expected. This page follows the whole category: coding agents and their IDE integrations, browser and computer-use agents, multi-agent orchestration frameworks, tool-calling standards like MCP, and the evaluation work that tries to establish which agents actually finish the tasks they start. Agent security gets its own attention here — prompt injection, tool poisoning and credential leakage — and we cover the incidents and the mitigations as they appear.
Quick Answers