Claude Code 2.1.220 focuses on reliability and terminal fixes
The latest Claude Code release continues the 2.1 line with reliability, terminal rendering, session, and developer-workflow fixes.
Source: official changelogCurrent and previous updates grouped by product. Open an agent to view its timeline, official changelog, and download link.
The latest Claude Code release continues the 2.1 line with reliability, terminal rendering, session, and developer-workflow fixes.
Source: official changelogClaude Code adds Claude Opus 5 support, strict allowlists, nested subagents up to depth three, directory hooks, and improved subagent forwarding.
Source: official changelogGitHub adds Claude Opus 5 to Copilot model choices across supported coding and agent experiences.
Source: official changelogGitHub Copilot cloud agents can be assigned and managed from Linear issue workflows with general availability support.
Source: official changelogQwen Code 0.19.12 through 0.20.1 add mid-run steering, one-click isolated worktrees, rebuilt web tools, and multimodal whole-turn routing.
Source: official changelogThe release improves background code review and ultrareview while tightening Windows behavior, accessibility, and security.
Source: official changelogCursor makes Auto an intelligent router with Intelligence, Balance, and Cost optimization modes plus administrator controls across product surfaces.
Source: official changelogCursor in Slack can plan before starting, work across multiple repositories, and continue tasks across channels.
Source: official changelogThe desktop app adds a global switcher for ChatGPT and Codex, unified recents and projects, while Codex keeps its separate view and existing history.
Source: official changelogGoogle turns Conductor into a plugin compatible with Antigravity CLI for persistent, specification-driven development workflows.
Source: official changelogQwen Code adds the /learn command, stronger code-review verification, a redesigned Web Shell, and voice-bridge integration.
Source: official changelogEligible organizations can configure HIPAA-ready controls through a self-serve Claude administration flow.
Source: official changelogClaude memory moves to individual categorized entries so users can inspect and manage retained preferences and context more precisely.
Source: official changelogCursor adds lightweight Side Chats for parallel questions and search across prior conversations.
Source: official changelogChatGPT Work adds a dedicated surface for longer tasks and moves browser-agent capabilities into ChatGPT and Codex as Atlas approaches retirement.
Source: official changelogThe release adds automatic model fallback chains, nested subagent spawning, Web Shell session management, and WeCom integration.
Source: official changelogCowork becomes available beyond desktop with remote sessions, scheduled tasks, and a shared home alongside Chat.
Source: official changelogGitHub expands Copilot agent-provider choices with Codex and ships related JetBrains agent improvements.
Source: official changelogTRAE App can minimize and resume voice conversations and add context with photos, images, and uploaded files.
Source: official changelogTRAE IDE adds a file-format subagent toggle and retries supported failures without losing pending tool calls.
Source: official changelogCodex Remote reaches general availability across ChatGPT plans and adds a DigitalOcean workspace plugin for cloud development workflows.
Source: official changelogTRAE Work launches Design Mode, improves voice chat with web and project context, and adds opt-in global memory.
Source: official changelogGemini CLI stopped serving free and individual Pro or Ultra accounts; Antigravity CLI becomes the unified multi-agent terminal path while enterprise and API-key access remains supported.
Source: official changelogCodex adds computer-use support on Windows, remote control improvements, and reusable profiles for coding environments.
Source: official changelogThe latest Claude Code release continues the rapid 2.1 line with reliability, terminal, permissions, plugin, and remote-control polish across developer workflows.
Source: official changelogAdmins can create and manage team marketplaces without first connecting a repository, and configure plugin distribution as optional, default-on, or required.
Source: official changelogCodex CLI adds persisted goal workflows, codex update, configurable keymaps, richer permission profiles, plugin marketplace flows, external-agent imports, and clearer MultiAgentV2 configuration.
Source: official changelogCursor introduces always-on security agents for pull-request review and scheduled vulnerability scanning, with configurable triggers, tools, and Slack reporting.
Source: official changelogxAI API responses now include exact request cost reporting across chat completions, Responses API, image generation, video generation, and streaming.
Source: official changelogChatGPT begins rolling out an optional Advanced Account Security setting with stronger sign-in methods, recovery keys, session controls, and tighter account safeguards.
Source: official changelogGoogle shuts down gemini-robotics-er-1.5-preview and points Gemini API users to gemini-robotics-er-1.6-preview.
Source: official changelogThe Cursor SDK lets developers build agents on the same runtime and harness used by Cursor desktop, CLI, and web agents, including local or cloud execution.
Source: official changelogClaude Code adds a non-interactive ultrareview command for CI, Windows PowerShell fallback, effort-aware skills, process attribution, plugin validation, and many session fixes.
Source: official changelogChatGPT makes model choice available from the composer and moves thinking-effort controls into the model picker for paid web users.
Source: official changelogCursor Agents Window gains multitask subagents, improved worktrees for isolated background branches, and multi-root workspaces for cross-repo changes.
Source: official changelogDeepSeek API adds V4-Pro and V4-Flash through OpenAI ChatCompletions and Anthropic interfaces, with legacy names scheduled for retirement.
Source: official changelogThe 0.123.0 release continues the CLI track with provider polish, MCP verbosity and loading improvements, remote sandbox configuration work, realtime handoffs, and model metadata updates.
Source: official changelogxAI makes grok-voice-think-fast-1.0 available for the Voice Agent API, expanding Grok voice-agent options for developers.
Source: official changelogGoogle releases gemini-embedding-2 as a generally available embedding model in the Gemini API.
Source: official changelogNew Gemini Deep Research previews add collaborative planning, visualization support, MCP server integration, and File Search for research workflows.
Source: official changelogThe CLI moves toward native per-platform binaries, adds domain deny controls, improves fullscreen editing, strengthens Bash safety, and fixes MCP concurrent-call handling.
Source: official changelogClaude Code adds Opus 4.7 xhigh effort, interactive effort controls, auto mode updates, less-permission-prompts, ultrareview, and broader Windows shell rollout.
Source: official changelogCursor agents can now create durable interactive canvases with tables, boxes, diagrams, charts, diffs, and to-do lists inside the agent side panel.
Source: official changelogGoogle launches Gemini 3.1 Flash TTS Preview as a cost-efficient, expressive, steerable text-to-speech model.
Source: official changelogThe updated Gemini Robotics ER model adds instrument reading plus improved spatial and physical reasoning capabilities.
Source: official changelogClaude Code adds worktree switching, PreCompact hook support, plugin background monitors, proactive aliasing, stalled stream fallback, clearer network errors, and doctor UI polish.
Source: official changelogQwen Code 0.14 adds multi-platform channels, cron scheduled tasks, /plan planning mode, adaptive output tokens, and Qwen 3.6 Plus model support.
Source: official changelogQwen3.6-Plus launches through Qwen research with a focus on real-world agents, agentic coding, long-context work, and multimodal reasoning.
Source: official changelogQwen3.5-Omni adds fully omnimodal understanding across text, images, audio, and audio-visual content with 256k long-context support.
Source: official changelogUpdated ChatGPT apps for Box, Notion, Linear, and Dropbox add new actions where supported and move those integrations to the newer app experience.
Source: official changelog