Google has introduced an experimental “Agent Mode” in the canary preview of Android Studio, allowing developers to delegate multi-step coding chores to Gemini, the company’s generative-AI model. The agent analyses a project, devises an execution plan that can span multiple files, and pauses for developer approval at key steps. The default Gemini model is available at no cost, while subscribers can tap Gemini 2.5 Pro’s one-million-token context window to handle larger codebases. Google says the feature, which communicates with external tools through the open Model Context Protocol, is designed to free developers from routine tasks while keeping them in control of code changes. The announcement comes a day after terminal-software startup Warp unveiled what it calls an “agentic development environment,” a workspace centred on prompt-driven collaboration with AI coding agents. Warp’s system lets programmers open multiple panes to supervise agents that generate, edit and deploy code, with safeguards that require human sign-off before critical actions. The company, which already has about 500,000 users, will keep its existing pricing—free, $15 and $40 a month—and says revenue is climbing 5 %–15 % a week. The two product launches underscore the quickening shift toward agentic AI, in which autonomous or semi-autonomous software agents perform complex tasks. Training providers are racing to meet demand: the Open Data Science Conference will host a three-week virtual Agentic AI Summit from 16–31 July, while Relevance AI and Data Science Dojo are opening new bootcamps and certification programmes aimed at helping engineers and business operators design, evaluate and scale multi-agent systems.