Articles, guides, research, and practical insights from Agent Zero.
v1.15 gives Markdown work its own first-class Editor surface, fixes unsafe chat markdown rendering, narrows Office and Desktop responsibilities, and improves connector and release-pipeline reliability.
v1.14 makes Agent Zero easier to start and cleaner to operate. The release adds guided first-run onboarding, brings Host Browser routing to the A0 CLI connector, separates Desktop and Office into focused plugins, and tightens tool, prompt, UI, documentation, and security behavior across the platform.
v1.13 makes Office workflows ODF-first, keeps the Linux Desktop lifecycle stable across canvases and modals, improves Browser switching and screenshot workflows, removes canvas sizing limits, and adds Bash-style chat input history.
v1.12 is a stability-focused release for sustained Agent Zero sessions. It closes PTY descriptor leaks, recovers cleanly from stale Browser contexts, routes Desktop URLs into the Browser, makes self-update backups resilient to broken symlinks, and starts the Desktop runtime only when it is needed.
v1.11 sharpens Agent Zero's browser into a stronger multi-tab workspace, replaces the Office stack with a LibreOffice-backed Desktop runtime, adds bulk operations to the file browser, simplifies navigation, and hardens the canvas, skills, Time Travel, and OAuth flows.
v1.10 introduces Agent Zero's new direct Playwright-powered Browser, ChatGPT/Codex account OAuth, Time Travel workspace history, Universal and Office Canvas surfaces, context-scoped Agent Profiles, project-scoped LLM presets, and a broad set of connector, UI, integration, and security improvements.
v1.9 closes two important security issues, adds the built-in A0 CLI Connector and its setup skill, restores lightweight lexical skill recall, gives Browser Agent runs their own model preset, and redesigns messaging integration setup flows across email, Telegram, and WhatsApp.
v1.8 makes skills easier to activate with a built-in selector directly in the chat composer, strengthens memory integrity and filtering safeguards, and restores reliable handling for tool calls that legitimately send no arguments.
v1.7 tightens default prompts to a compact assembled stack while preserving guardrails, dispatches tool calls as soon as a complete JSON object appears in the model stream, and adds discovery cards on the welcome experience so new users can find the Plugin Hub and messaging integrations quickly.
v1.6 introduces a powerful new WhatsApp integration plugin, major UI server restructuring into modular components, Settings enhancements ensuring API keys aren't overwritten, and update system improvements.
v1.5 ships a deep WebSocket architecture rewrite with stronger routing and buffering primitives, lower initialization complexity, and multiple security and memory-leak fixes. It also includes model prompt optimization, upload-path robustness, and UI/installer quality improvements.
v1.4 focuses on first-run usability and plugin ecosystem quality. It adds API-key onboarding guidance, improves model preset UX in chat, and hardens markdown rendering pipelines in plugin documentation views.