Jules, remote coding agent from Google Labs, announces API (jules.google)

Google Labs has opened programmatic access to Jules, its remote coding agent, with a new Jules API and a command-line tool, Jules Tools. Developers can now integrate Jules into ChatOps (Slack), project trackers (Linear, Jira), and CI/CD (GitHub Actions), create sessions via cURL, and script interactions with commands like jules remote new and jules remote list. The CLI is installable via npm (npm install -g @google/jules) or runnable with npx, and supports a TUI, piping with tools like gh/jq, and an “apply patches locally” flow to pull work-in-progress code onto your machine without waiting for a GitHub commit. Beyond connectivity, Jules is gaining production-ready features that matter technically: per-repository Memory (toggleable) to persist preferences and corrections; precise file selection; PR-feedback automation with an optional Reactive Mode; image upload at task creation (JPEG/PNG, total <5MB); Playwright in the base VM for frontend testing and screenshots; a 20GB VM to handle large repos; and a critic agent whose step-by-step analysis is now visible. Jules can also web-search technical docs, render images inside diffs, publish WIP branches/prs at any time, and uses Gemini 2.5 for planning. For ML/AI practitioners, this marks a step toward composable, auditable coding agents that integrate into developer pipelines—boosting automation and reproducibility but raising implications for repo-level privacy, memory governance, and CI security.