AI pair programmers and autonomous coding agents — IDE-integrated assistants and cloud agents that read, edit, and ship code.
OpenAI brought GPT-6.1 Sol Ultrafast to Codex as its new default, Anthropic hardened Claude Code's sandbox and shipped a cheaper Haiku default, and Cursor let developers steer their local agents from an iPhone.
Devin now carries personal memory across sessions with a nightly review pass and org-level controls, while cutting its session resume window from 30 days to 14.
Origin now integrates with over 20 apps for deployments, CI, code review, security, monitoring, and project management, from partners including Cloudflare, Linear, Sentry, and Socket Security. The marketplace is open to public submissions, with apps authenticating via app JWT and installation access tokens and writing back check runs, pull request comments, and reviews through the Origin API.
Devin now remembers what it learns about you across sessions, with memory files reviewable and toggleable under Settings > Customization > Memory and a nightly review of recent sessions listed as recent dreaming sessions. Sessions can now be resumed for 14 days after going to sleep, down from 30. Plus a long tail of updates: font and font size settings, command palette shortcuts, real transcript diff cards for large edits, per-type notification switches, server-synced user preferences, AGENTS.md loaded on repo setup, and non-expiring PR attachment links on Enterprise.
xAI1Shell commands no longer keep running after quitting grok or after grok receives a signal, and grok worktree rm and gc no longer delete worktrees that open sessions are using or directories grok did not create. Fixed a model switch reverting on restart, automatic compaction on context overflow errors and after repeated refills, and MCP replies with invalid characters timing out, plus smaller fixes across hooks, /model, and the terminal UI.
Claude Code stopped Edit and NotebookEdit from mangling non-UTF-8 files and made blocking managed hooks actually refuse tool calls, alongside new subagent compaction and gateway policy controls.
Managed-settings PreToolUse and prompt hooks that block a tool call now refuse it and end the turn, and Edit and NotebookEdit no longer replace every non-ASCII character in files that are not valid UTF-8 — such edits are refused instead. Adds a code key to the Claude apps gateway's managed.policies[], autoCompactWindow for subagents, CLAUDE_CODE_WORKFLOW_SUBAGENT_MODEL, an allow_large option for the Read tool, plus a long list of fixes across MCP servers, plugins, self-hosted runners, cloud sessions, and Windows.
Claude Code hooks can now fail closed with an onFailure block, gateway upstreams accept per-endpoint model allowlists, and instruction-style prompt and agent hooks no longer let blocked actions through.
xAI1The /share command and session sharing are removed from the CLI, and memory is now only an on or off switch. Cancel and edit-resend on long sessions no longer revert workspace files, and headless grok -p waits for background goals to finish and reports their outcome.
Fixed a TUI crash when asynchronous questions contain multiple lines, now preserving line breaks and complete hyperlink destinations. Startup failures caused by feature-setting differences between a running background server and CLI defaults are also fixed, with compatibility checks now applying only to explicit command-line feature overrides.
Command and HTTP hooks support onFailure: "block" so a hook that fails to start, times out, or exits unexpectedly blocks the action instead of allowing it, and Claude apps gateway upstreams now accept an optional models list that limits which models are sent there, including on failover. Also fixed context-1m beta rejection breaking all requests on [1m] models, remote MCP reconnects looping against servers that drop connections, and claude -p text output dropping earlier responses when background work started another turn.
Fixed an issue where prompt and agent hooks written as instructions, such as "Block commands that...", allowed what they should block. Improved how prompt hooks on Stop and SubagentStop written as instructions are judged, so Claude is less likely to stop early.
Claude Code made Claude Haiku 5.5 its default Haiku model with 1M context at lower per-token pricing and fixed a memory leak in HTTP MCP connections, while Devin shipped a notification inbox with a settings page for inbox and phone delivery.
Adds tools for creating and listing managed Git worktrees from trusted local projects when the worktrees feature is enabled, plus /copy for navigating and copying transcript blocks, Ctrl+Insert to copy selections, and tui.mouse_scroll_speed for mouse-wheel scrolling. Fixed Linux sandbox startup with multiple denied files and restored ordinary drive-letter file access on Windows 10.
Ultrafast mode speeds up token generation with GPT-6.1 Sol in Codex and ChatGPT Work, available on Pro $500 and eligible Enterprise and Edu plans. Enterprise access is off by default and workspace owners must enable it; inference residency is supported in the United States and Europe (EEA + Switzerland).
Devin now has a notification inbox with a bell in the sidebar and a Settings > Notifications page for choosing inbox and phone delivery, the ⌘K palette mirrors sidebar pages and exposes pull request review actions, and saved secrets containing a private key are redacted line by line in command output. Other changes cover Microsoft Teams and Slack messaging, merge queue status, Devin Review permissions for custom roles, security profiles for Teams group chats, analytics corrections, and plugin row and Settings dialog styling.
Claude Haiku 5.5 (claude-haiku-5-5) is now the default Haiku model on the Anthropic API with 1M context and $0.10/$0.50 per Mtok. Fixed a memory leak where an HTTP MCP connection retained every request until it closed, along with fixes for work retraction after context compaction and lost messages when backgrounding sessions.
Claude Code patched multiple sandbox and permission bypasses — including sandboxed commands reading staged review uploads and a tampered settings cache disabling the policy plugin — while adding marketplace-sourced plugin installs and an effort parameter for sub-agents.
GPT-6.1 Sol is now the default model in the bundled and Amazon Bedrock catalogs, and Daybreak is now opt-in via --enable cli_daybreak or features.cli_daybreak=true, with controls, indicators, and /daybreak hidden and automatic Cyber routing omitted by default. Also added /mcp login <name> for signing in to MCP servers, per-turn Cyber access program selection via codex exec --cyber-access-program, and voice input/output device selection with locally saved preferences.
Page previews now render inline in responses, and Codex task links can be opened directly on iOS. The approval picker is redesigned with clearer permission choices, plugins are reachable from a single Plugins entry in the composer's plus menu, and worktree support now matches the checkout used on desktop.
Claude Code v2.1.292 adds a --marketplace <source> option to claude plugin install and an effort parameter for the Agent tool, plus an overloaded-request retry delay env var. It also closes several sandbox and permission bypasses, including sandboxed commands reading staged /ultrareview upload file copies, managed read-deny paths not dropping project grants, network-path reads skipping the permission prompt, and a tampered settings cache disabling the policy plugin.
Cognition's Devin gained an in-session Terminal tab with keyboard shortcuts and fixed long-session transcript flicker and Teams access enforcement, while Claude Code widened its mod hook API and patched proxy, image, and compaction failures.
Fixed a 2.1.290 regression where cloud sessions could drop answers to permission prompts, and a 2.1.288 regression where a session's final messages could be lost when quitting.
Mod and plugin hooks gain richer context: serverToolUses in turn.step results, agentId on tool.check events, a ceiling on required approvals, and theme typing for drawing. The release also fixes requests failing behind proxies that reject beta headers, sessions with hundreds of images getting stuck on unprocessable-request errors, WebFetch silently dropping text past 100,000 characters, and compaction failures plus scheduled tasks vanishing on resume. Added claude attach and claude logs accepting partial session names, and a Deny button on the Claude apps gateway sign-in approval page.
The session workspace gains a Terminal tab for running commands on Devin's machine, with Files and Terminal now waking a sleeping session, and the Settings keyboard shortcuts page is available to everyone. Also adds single code scan fetch via API and MCP, Bitbucket Data Center PR reviewer assignment, and Azure DevOps review summary threads posted as closed.
xAI1grok worktree rm without -f now refuses to delete worktrees that are in use or have unsaved changes and exits with an error, and long sessions stay connected when the conversation exceeds 4 MiB. The release also retries stalled model calls and Chat Completions streams that end without a finish reason, prevents concurrent settings edits from dropping config.toml entries, and adds a timestamps setting plus TSV/CSV/Markdown table export.
Cursor's iOS app can now display and message local agents running on your computer, with computers pairing to the account automatically and agents continuing to run locally. Remote Control is on by default except for Enterprise organizations, where admins enable it under Org settings > Security and identity > Remote control.
Codex
Claude Code
Cursor
Devin
Grok Build