OpenCode v1.0 Released (opencode.ai)

OpenCode v1.0 is a terminal-first AI coding agent that lets developers bring LLM-powered assistance directly into their shell. It’s distributed via an install script, platform packages (Node.js, Homebrew, Paru, Chocolatey, Scoop, NPM) or downloadable binaries (Windows/Bun support still in progress). OpenCode is model-agnostic: you configure API keys for any LLM provider (the project recommends using its curated OpenCode Zen model list), authenticating with opencode auth login → opencode.ai/auth. Once set up, OpenCode can analyze a repository and generate an AGENTS.md in the project root to capture structure and coding patterns. The tool emphasizes an iterative, auditable workflow: initialize a project to let the agent inspect files, use Plan mode (Tab) to get a change plan without modifying code, then switch to Build mode to apply changes. It supports dragging images into the terminal to include visual context, and provides /undo and /redo commands to revert or reapply edits. Conversations are shareable via generated links for team review. For the AI/ML community this means a reproducible, model-agnostic developer assistant that integrates with existing CI and code-review practices, encourages plan-before-change safety, and lowers friction for applying multimodal models in real coding workflows — though teams should still manage API keys, billing, and review outputs for security and correctness.