Updated daily

Daily updates

Every release, changelog entry and community signal for Claude Code, Codex, Cursor and Grok Build - collected daily, grouped by day. The raw feed the weekly brief is curated from. Prefer it curated? Read the weekly brief.

By day

8

Claude CodeReleaseanthropics

Claude Code v2.1.175 — enforceAvailableModels managed setting

Added `enforceAvailableModels` managed setting — when enabled, the `availableModels` allowlist also constrains the Default model (a Default that would resolve to a disallowed model now falls back to the first allowed model), and user or project settings can no longer widen a managed `availableModels` list.

Claude CodeReleaseanthropics

Claude Code v2.1.174 — wheelScrollAcceleration setting, /model picker fixes, VSCode usage attribution

Added wheelScrollAccelerationEnabled setting to disable mouse-wheel scroll acceleration in fullscreen mode. Fixed the /model picker hiding the model family that Default resolves to. Fixed /model picker showing hardcoded Sonnet version label. Fixed Fable 5 banner for enterprise accounts. Fixed Bedrock GovCloud regions deriving wrong inference profile prefix. Fixed background sessions inheriting provider env from shell. Fixed 1-2 second pause when exiting Claude Code on macOS/Linux. Fixed git commit co-author attribution. Fixed /advisor dialog pre-selecting blocked model. Fixed skill hot-reload re-sending entire listing. [VSCode] Added usage attribution to Account and usage dialog with cache misses, long context, subagents, and per-skill/agent/plugin/MCP breakdowns. Fixed pre-warmed background workers failing with authentication error.

Claude CodeReleaseanthropics

Claude Agent SDK TypeScript v0.3.174 — model_fallback trigger field expanded

SDK consumers now receive the system/model_fallback message for all fallback triggers — overloaded, server_error, and last_resort in addition to model_not_found and permission_denied — and the message trigger field gained the server_error and last_resort values. Install: npm install @anthropic-ai/claude-agent-sdk@0.3.174

12

Claude CodeNewsMarco Quiroz-Gutierrez / Fortune

Anthropic reverses: AI will now tell users when requests are rejected or downgraded for national security

Anthropic has reversed course on transparency regarding its Fable 5 model after facing criticism from the AI research community. The company initially buried disclosure in a 319-page safety document that its advanced AI system would silently downgrade requests related to frontier AI development. Following backlash, Anthropic committed to making these restrictions visible. Users will now receive notifications when their requests are downgraded or rejected. According to an Anthropic spokesperson: Flagged requests will visibly fall back to Opus 4.8... any flagged requests will return a reason for their refusal. The company continues downgrading certain requests for two reasons: preventing use of its models to build competing AI systems (a standard industry practice) and protecting national security. Anthropic stated: The U.S. and its allies hold an edge in frontier chips... These safeguards ensure Claude is not used to erode that advantage. Researchers like Jeremy Howard criticized the initial approach, arguing it would slow AI development. Anthropic acknowledged the misstep, noting: We made the wrong tradeoff and we apologize for not getting the balance right.

Claude CodeNewsRichard L. Wells / TechTimes

AI Regulation Push: Amodei Demands Government Power to Block Unsafe Models, Anthropic Pledges $350M

Anthropic CEO Dario Amodei released a comprehensive policy proposal demanding that the U.S. government gain authority to block or reverse frontier AI model releases that fail independent safety testing. Alongside this essay, Anthropic announced $350 million in funding commitments. The Advanced AI Framework targets models trained with over 10^25 floating-point operations from companies with substantial AI revenue or research spending. Qualifying systems would undergo third-party testing in four risk domains: biological weapons, cybersecurity, loss of control, and autonomous AI development. The Economic Policy Framework proposes scaling responses to three disruption scenarios, from modest unemployment increases to unprecedented labor market shifts, suggesting wage insurance, retraining programs, and potentially universal basic income at higher disruption levels. The announcement drew mixed reactions — supporters cite documented risks, while critics worry the framework could entrench Anthropic market position while raising barriers for competitors.

Claude CodeReleaseanthropics

Claude Code v2.1.173 — Fable 5 [1m] suffix normalization fix

Fixed Fable 5 model names with a [1m] suffix not being normalized — Fable 5 includes 1M context by default, so the suffix is now stripped automatically. Fixed spurious sandbox dependencies missing startup warning on Windows when sandbox was enabled in settings.

Claude CodeNewsDiane Brady / Fortune

Fortune: What Anthropic's Fable 5 means for CEOs trying to govern AI — data retention, silent limits, access tiers

Fortune's Diane Brady examines the challenges Anthropic's Claude Fable 5 release presents for corporate leadership, capturing widespread concern about escalating complexity in AI governance. Three primary concerns identified: 1. Capability Restrictions: Fable 5 automatically blocks responses in cybersecurity and limits AI researcher capabilities, with an unrestricted version available only to preapproved organizations. 2. Transparency Issues: The model silently limits outputs without alerting users, creating uncertainty about when they receive reduced functionality. 3. Data Privacy: Unlike earlier Anthropic offerings with zero data retention, Fable 5 maintains a 30-day retention window, prompting firms like Microsoft to restrict employee access. Bezalel Eithan Raviv, a cybersecurity CEO, argues that Anthropic has officially become the judge and the executioner regarding safety standards, contending these models demand regulatory oversight similar to currency or weapons manufacturing.

Claude CodeNewsMarco Quiroz-Gutierrez / Fortune

Anthropic reverses Fable 5 hidden restrictions, commits to transparency on safeguard fallbacks

After backlash over silent capability downgrades in Claude Fable 5, Anthropic announced on June 11 that the model will now explicitly notify users when a request is being rejected or rerouted due to national-security safeguards. Flagged requests will display a fallback notification; API refusals will include explanatory reasons. Anthropic will continue applying restrictions related to frontier AI development and adversarial capability uplift, but the shift prioritizes transparency over silent content filtering. (Marco Quiroz-Gutierrez / Fortune, June 11 2026)

Grok BuildNewsSarah Chen / Basenor

xAI launches Grok Build Plugin Marketplace with MongoDB, Vercel, Sentry, Chrome DevTools, Cloudflare

xAI unveiled the Grok Build Plugin Marketplace on June 11, 2026, with Sentry integration as its inaugural offering. This plugin enables developers to ask their AI agent to find and fix errors, analyze stack traces, and triage production alerts directly from the terminal. The Plugin Marketplace is built into Grok Build, so you can browse, install, and update plugins without leaving your terminal. A plugin bundles skills, slash commands, agents, hooks, MCP servers, and LSPs into one installable package. Launch catalog ships six plugins: MongoDB (explore data, manage collections, optimize queries), Vercel (manage deployments, check build status, configure domains), Sentry (analyze stack traces and debug production errors), Chrome DevTools (control a live browser, record performance traces, inspect network requests), Cloudflare, and Superpowers. Currently limited to SuperGrok and X Premium Plus subscribers. Every remote plugin in the catalog is pinned to a specific commit SHA, and Grok Build verifies the pin at install time. The marketplace is open — submit a pull request to xai-org/plugin-marketplace to add your plugin. Grok Build operates at .00 per million input tokens and .00 per…

16

Claude CodeReleaseanthropics

Claude Code v2.1.172 — sub-agents can spawn sub-agents (5 levels deep), Bedrock AWS config

Sub-agents can now spawn their own sub-agents (up to 5 levels deep). Amazon Bedrock now reads AWS region from ~/.aws config files when AWS_REGION is not set. Added search bar when browsing marketplace plugins in /plugin. Added model attribute to claude_code.lines_of_code.count OTEL metric. Fixed sessions using 1M context without usage credits getting stuck. Fixed repeating image-in-conversation error. Multiple agent panel, /model picker, and Bedrock fixes. Disable mouse tracking on Windows consoles. Improved performance in long conversations. Reduced idle CPU usage. Improved Claude in Chrome tool loading. [VSCode] Fixed PowerShell tool calls rendering.

Claude CodeReleaseanthropics

Claude Agent SDK TypeScript v0.3.172 — skipMcpDiscovery per plugin option

SDK plugins option now accepts skipMcpDiscovery: true per plugin, so a host that manages a plugin MCP connections itself can load skills/hooks from the plugin path without the engine re-reading its .mcp.json. Fixed slash-followed-by-whitespace input being silently dropped. Install: npm install @anthropic-ai/claude-agent-sdk@0.3.172

Claude CodeNewsSharon Goldman / Fortune

Anthropic walks back covert capability limits on Claude Fable 5 after 'secret sabotage' accusations

Hours after Fable 5 launched on June 9, researchers discovered a paragraph buried in its 319-page system card revealing the model silently degrades responses for certain AI research tasks — without informing users — to limit frontier LLM development by adversaries. Critics called it "secret sabotage," arguing Anthropic was using safety justifications to consolidate competitive advantage while retaining the capability for its own use. Facing intense social-media backlash, Anthropic reversed course on June 10, stating: "We made the wrong tradeoff, and we apologize for not getting the balance right." The company committed to making all such restrictions visible to users going forward. (Sharon Goldman / Fortune, June 10 2026)

Claude CodeNewsSharon Goldman / Fortune

Anthropic accused of 'secret sabotage' as Claude Fable 5 silently limits capabilities for AI researchers

Anthropic reversed a policy that had quietly restricted its new Claude Fable 5 model's ability to assist with AI research tasks. The company initially released the Mythos-tier model with safeguards that would downgrade responses related to frontier AI development without telling users, as stated in the 319-page system card. The backlash was swift and severe. Researchers complained that receiving diminished answers without notification violated user expectations. Nathan Lambert wrote that the covert approach was appalling, while Dean Ball criticized the 'secret sabotage' safety policy as potentially proving AI safety concerns were merely hype to justify monopolistic behavior. Even former Anthropic employee Behnam Neyshabur expressed concern, noting the restrictions would slow scientific and technological progress. In response, Anthropic acknowledged the misstep. A company spokesperson stated: We made the wrong tradeoff, and we apologize for not getting the balance right. The firm committed to making these restrictions visible to users going forward, similar to other safeguards already in place.

Claude CodeNewsSwati Khandelwal / The Hacker News

Anthropic Releases Claude Fable 5 — two-tier model: Fable (public) and Mythos 5 (restricted to defenders)

Anthropic released Claude Fable 5 on June 9, 2026, alongside a restricted variant called Claude Mythos 5. The company deployed an unusual distribution model: the same underlying model offered as two products, differentiated by safety controls rather than capability. - Fable 5 (public): Routes flagged cybersecurity, biology, chemistry, and model-distillation requests to the weaker Claude Opus 4.8 - Mythos 5 (restricted): Maintains full cybersecurity capabilities for vetted defenders and infrastructure operators Separate AI classifiers monitor for misuse and jailbreak attempts. When triggered, the system hands requests to the downgraded model while notifying users. Testing showed the safeguards blocked progress on harmful tasks, with fewer than 5% of sessions triggering fallback mechanisms. Pricing: Both models cost 0 per million input tokens and 0 per million output tokens, less than half the earlier Mythos Preview pricing. Anthropic implemented a 30-day retention requirement for all traffic on these capability levels, with no training usage permitted.

CursorNewsCursor

Cursor Bugbot: 3x faster (90s avg), 22% cheaper, 10% more bugs found — plus new /review command

Cursor shipped major Bugbot improvements on June 10 2026. Average review time dropped from ~5 minutes to ~90 seconds (3x faster), bugs found per run increased from 0.56 to 0.62 (+10%), and per-run cost fell ~22%. The gains come from Composer 2.5 now powering Bugbot. New: run `/review` (or `/review-bugbot`, `/review-security`) before pushing to get feedback pre-PR. `/review` syncs with Bugbot on GitHub/GitLab — if you open a PR with the same diff, Bugbot skips a redundant review. Also new: configurable incremental reviews — Bugbot can now be set to only review changes since the last review.

Claude CodeNewsThomas Claburn / The Register

It blocked us at 'hello!' — Anthropic Fable 5 refusing innocuous prompts, company reverses course

Anthropic's newly released Claude Fable 5 model is blocking legitimate user requests due to overly aggressive safety classifiers. The company acknowledged using conservative guardrails that trigger on harmless inputs, though stated they occur in less than five percent of sessions. Users reported significant frustrations across multiple platforms. A researcher at the Gates Foundation noted the model refused basic inputs like 'hello,' while an immunologist flagged that terms related to biology research trigger false positives. Additional complaints documented refusals for resume editing and legitimate research tasks. The system employs hidden safeguards that secretly degrade responses when detecting AI development or security work — what one developer characterized as purposeful degradation invisible to the user. Following backlash, Anthropic issued a statement accepting responsibility. The company committed to making safeguards transparent, with visible fallbacks to Opus 4.8 and API explanations for flagged requests. Leadership acknowledged making the wrong tradeoff and committed to reducing false positives while maintaining protections against frontier model development.

Claude CodeNewsJose Antonio Lanz / Decrypt

The Internet Is Furious at Anthropic After Claude Fable 5 Release — token burn, silent degradation, GDPR data retention

The community reacted negatively to Anthropic Fable 5 release across three major issues. 1. Token Consumption: Users reported the model drains subscription allowances rapidly. One test showed a $100 Max plan depleted in under nine minutes. A CEO calculated usage equivalent to $333k/year salary based on token burn rates, while developers described spending over $1,000 daily on limited plans. 2. Silent Performance Degradation: Anthropic disclosed that Fable 5 covertly reduces its own capabilities for AI research topics without user notification. One researcher stated: It would be nice not to ban biomedical scientists before talking access. Critics argued this undermines scientific reproducibility and transparency. 3. Data Retention Policies: The mandatory 30-day data retention requirement locked out European companies requiring zero-retention compliance under GDPR, prompting concerns about enterprise usability and regulatory alignment. Developers, researchers, and open-source advocates expressed frustration across social media, with many declaring they would discontinue using Anthropic services.

Claude CodeNewsDario Amodei / Anthropic

Dario Amodei: 'Policy on the AI Exponential' — aviation-style regulation, government power to block unsafe models

Anthropic CEO Dario Amodei published a comprehensive policy essay on June 10, 2026, arguing that the US government should hold legal authority to block or reverse the release of frontier AI models that fail independent safety testing. Key proposals: 1. Safety Regulation: Mandatory third-party testing of advanced models for risks in cybersecurity, biological weapons, loss of control, and automated R&D. Governments should possess authority to block deployment of models presenting unacceptable risks. 2. Economic Policy: Recognizing AI may cause significant job displacement, Amodei proposes wage insurance, training grants, and potentially universal basic income financed through corporate or capital gains taxation. 3. Scientific Acceleration: Regulatory agencies should preemptively develop standards for AI-based drug development methods. 4. Civil Liberties: Establish accountability mechanisms for autonomous weapons systems, ban such weapons for domestic law enforcement, close privacy loopholes exploited by bulk data collection. 5. Geopolitics: Democracies should coordinate as a coalition managing critical AI supply chains. Alongside the essay, Anthropic pledged $350 million: a $200 mil…

CursorChangelogCursor

Cursor Bugbot: 3x faster (90s reviews), 22% cheaper, 10% more bugs found

Cursor announced significant performance improvements to Bugbot, its code review tool. The average review time for Bugbot is now ~90 seconds, down from ~5 minutes, with bug detection increasing to 0.62 bugs per review from 0.56, while costs decreased by approximately 22%. The enhancements leverage Composer 2.5, the underlying model powering Bugbot. New features include the ability to run code reviews locally using the /review command before pushing changes. This integrates with GitHub and GitLab, allowing Bugbot to recognize previously reviewed diffs and skip redundant analysis. Available in Cursor 3.7 and later. Users can now configure Bugbot to review only newly changed code since the last review, focusing feedback on recent modifications rather than the entire codebase.

35

Claude CodeReleaseanthropics

Anthropic TypeScript SDK v0.104.1

## @anthropic-ai/sdk v0.104.1 Released June 9, 2026 at 23:55 UTC. ### Bug Fixes - Added `frontier_llm` refusal category Install: `npm install @anthropic-ai/sdk@0.104.1`

CodexReleaseopenai

Codex CLI rust-v0.139.0 — web search in code mode, oneOf/allOf schema preservation

New features: - Code mode now supports standalone web search from nested JavaScript tool calls with plaintext results - Tool schemas preserve oneOf/allOf structures; large schemas maintain shallow structure for better MCP compatibility - codex doctor includes editor and pager environment details while redacting values in JSON Bug fixes: - resume and fork commands now correctly interpret trailing arguments as initial prompts - MCP startup warnings from subagents remain in their own thread, preventing duplicate alerts - Image edits use exact referenced file paths instead of guessing from conversation history - Bare URLs with tildes are now fully linkified in the TUI - Thread resets preserve cloud-managed requirements and feature flags Chores: - Release builds publish separate symbol archives with line tables for crash symbolication - Embedded V8 toolchain updated to rusty_v8 149.2.0

Claude CodeReleaseanthropics

Anthropic TypeScript SDK v0.104.0

## @anthropic-ai/sdk v0.104.0 Released June 9, 2026 at 20:04 UTC. ### Features - Added support for Managed Agents deployments and environment variable credentials Install: `npm install @anthropic-ai/sdk@0.104.0`

Claude CodeFrom X@bcherny

Boris Cherny: Fable 5 stepped up from coding agent to thought and design partner

Fable 5 is the biggest step up I have felt in our models since Opus 4.5 back in November. After 4.5 came out I uninstalled my IDE when I realized that I had been doing 100% of my coding in a terminal for a few weeks. With Fable, it has felt like Claude has stepped up from being a coding agent to a thought and design partner in building the product. Fable has judgement, taste, and dimensionality in a way that previous models did not, leading me to trust it more with the most complex work.

Claude CodeFrom X@karpathy

Karpathy on Fable 5: major-version-bump step change, most tempting to stop looking at code entirely

This is a super exciting release - Claude Fable 5 is the same underlying model as Mythos but with added safeguards. The benchmarks are great and it is SOTA on everything by a margin but I will add that qualitatively also, this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems. You can give it a lot more ambitious tasks than what you are used to, the model gets it and it will just go, and it has never felt this tempting to stop looking at the code at all.

Claude CodeFrom X@bcherny

Boris Cherny: Fable 5 available in Claude Code — best coding model by wide margin

Fable 5 is now available in Claude Code and Cowork Fable is the best model I have used for coding, by a wide margin. It is a big step up, enabling less prompts and steers, more efficient token use, better code quality, better tool use, more intelligent self-verification, longer running sessions, and higher trust and autonomy. Happy coding!

Claude CodeReleaseanthropics

Claude Agent SDK (Python) v0.2.95

## claude-agent-sdk 0.2.95 Released June 9, 2026 at 17:40 UTC. - Updated bundled Claude CLI to version 2.1.170 Install: `pip install claude-agent-sdk==0.2.95`

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.170

## What's changed - Added claude-fable-5 model and the fable alias to SDK model types. https://www.anthropic.com/news/claude-fable-5-mythos-5 - Updated to parity with Claude Code v2.1.170 ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.170 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.170 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.170 # or bun add @anthropic-ai/claude-agent-sdk@0.3.170 ```

Claude CodeReleaseanthropics/claude-code

v2.1.170

## What's changed - Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use. Fable’s capabilities exceed those of any model we’ve ever made generally available. Update to version 2.1.170 for access. https://www.anthropic.com/news/claude-fable-5-mythos-5 - Fixed sessions not saving transcripts (and not appearing in --resume) when launched from the VS Code integrated terminal or any shell that inherited Claude Code environment variables.

Claude CodeFrom X@claudeai

Fable 5 safety: dangerous capability queries fall back to Opus 4.8

Releasing a model this capable comes with risks. Without safeguards, Fable 5 capabilities in areas like cybersecurity could be misused to cause serious damage. Queries on a narrow range of topics will instead receive a response from our next-most-capable model, Opus 4.8.

Claude CodeNewsRebecca Bellan / TechCrunch

Anthropic releases Claude Fable 5, its most powerful model publicly, days after warning AI is getting too dangerous

Anthropic released Claude Fable 5 on June 9, 2026, making its powerful Mythos-class model publicly accessible for the first time through the Claude API and Enterprise plans. Subscription access rolls out in stages, with free inclusion through June 22, then requiring usage credits. The model excels at software engineering, knowledge work, and vision tasks. Third-party testing showed Fable was the first to get a 90% on its core analytics benchmark of complex, long-running analytical tasks. Fable includes hard safety limits that block responses in high-risk areas including cybersecurity, biology, chemistry, and model distillation. In these cases, the system defers to Claude Opus 4.8. Early data indicates at least 95% of Fable sessions run entirely on Fable own responses. Both Fable 5 and Mythos 5 cost 0 per million input tokens and 0 per million output tokens — double the price of Opus 4.8. Anthropic conducted extensive stress-testing with jailbreak attempts and external red-teaming, finding no universal jailbreaks in over 1,000 hours of testing.

Claude CodeReleaseanthropics

Anthropic Bedrock SDK v0.30.2

## @anthropic-ai/bedrock-sdk v0.30.2 Released June 9, 2026 at 16:36 UTC. ### Bug Fixes - Fixed 3p middleware ordering (#53) Install: `npm install @anthropic-ai/bedrock-sdk@0.30.2`

Claude CodeReleaseanthropics

Anthropic TypeScript SDK v0.103.0 — Fable 5 & Mythos 5 support

## @anthropic-ai/sdk v0.103.0 Released June 9, 2026 at 16:36 UTC. ### Features - Added support for `claude-mythos-5` and `claude-fable-5` with server-side fallbacks on refusal - Client-side fallbacks middleware for API providers without server-side fallback support - Added `ctx.logger` to middleware (#55) ### Bug Fixes - Fixed 3p middleware ordering (#53) Install: `npm install @anthropic-ai/sdk@0.103.0`

Claude CodeReleaseanthropics

Anthropic Vertex SDK v0.17.1

## @anthropic-ai/vertex-sdk v0.17.1 Released June 9, 2026 at 16:36 UTC. ### Bug Fixes - Fixed 3p middleware ordering (#53) Install: `npm install @anthropic-ai/vertex-sdk@0.17.1`

Claude CodeNewsCNBC

CNBC: Anthropic releases Mythos-like AI model to the public as Claude Fable 5

Anthropic launched Claude Fable 5, the first publicly available version of its Mythos model, on June 9, 2026. Claude Fable 5 is a Mythos-class model that has been made safe for general use, with capabilities that exceed those of any model Anthropic has ever made generally available. The model shows exceptional performance in software engineering, knowledge work, vision, scientific research, and many other areas. During early testing, Stripe reported that Fable 5 compressed months of engineering into days, performing a codebase-wide migration in a 50-million-line Ruby codebase in a day that would otherwise have taken a whole team over two months. The model was launched with safeguards that mean queries on some topics will receive a response from Claude Opus 4.8 instead, with safeguards triggering in less than 5% of sessions on average. From June 9 through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost. Fable 5 and Mythos 5 are being offered at $10 per million input tokens and $50 per million output tokens.

Claude CodeNewsAnthropic

Anthropic releases Claude Fable 5 — first public Mythos-class model

Anthropic launched Claude Fable 5 on June 9, 2026 — its first publicly available 'Mythos-class' model, a tier above Opus 4.8 (priced ~$10/$50 per Mtok in/out). It targets long-horizon agentic work: codebase migrations, multi-hour builds, agent loops, vision-heavy tasks, and research synthesis. Anthropic's Lance Martin demonstrated Fable 5 self-correcting in rubric-driven loops far more effectively than the prior generation.

Claude CodeFrom XClaudeDevs

ClaudeDevs: Claude Fable 5 is here — new model generation, new way of working

Claude Fable 5 is here. New model generation, new way of working. Here's how to get started in Claude Code and on the Claude Platform. In the API, switch the model ID parameter to "claude-fable-5" in both the Messages API and Claude Managed Agents. For some accounts, an admin may need to enable access first. The /claude-api skill in Claude Code can help you build with Fable on the API.

Claude CodeNewsRebecca Bellan / TechCrunch

Anthropic's Claude Fable 5 is a version of Mythos the public can access today

TechCrunch coverage of the Claude Fable 5 and Mythos 5 launch (Rebecca Bellan, June 9 2026). Fable 5 is accessible via the Anthropic API and enterprise plans, with free access for subscribers through June 22. Anthropic conducted a bug bounty program and over 1,000 hours of red-teaming, reporting no universal jailbreaks found. A 30-day traffic retention policy is now required for all users to defend against novel attacks.

Claude CodeNewsAnthropic

Introducing Claude Fable 5 and Claude Mythos 5

Anthropic released Claude Fable 5, a Mythos-class model made safe for general use, on June 9 2026. Its capabilities exceed any model Anthropic has previously made generally available, with state-of-the-art performance on nearly all benchmarks — especially for long, complex tasks in software engineering, knowledge work, vision, and scientific research. Fable 5 includes safety classifiers that automatically fall back to Claude Opus 4.8 for queries flagged as high-risk (cybersecurity, biology/chemistry, model distillation). These safeguards trigger in fewer than 5% of sessions on average. Claude Mythos 5, the same underlying model with certain safeguards lifted, is available only to authorized cyber-defenders and infrastructure providers via Project Glasswing. Pricing: $10/M input tokens, $50/M output tokens. Free on Pro/Max/Team/Enterprise plans through June 22; usage credits required after that.

Claude CodeNewsSimon Willison

Simon Willison: Initial impressions of Claude Fable 5 — a beast, expensive, successful on every task

Simon Willison spent approximately 5.5 hours testing Claude Fable 5 and characterized it as something of a beast. The model is powerful yet slow and expensive, successfully handling every task he attempted. The model features a 1 million token context window with 128,000 maximum output tokens and a January 2026 knowledge cutoff. Pricing is double that of Claude Opus 4.5-4.8: 0 per million input tokens and 0 per million output tokens. Fable demonstrated substantially greater knowledge depth than Opus 4.8. Willison used it to upgrade his micropython-wasm project to full Python support and to implement human-approval features in Datasette Agent, resulting in a complete library release essentially authored by the model. After a full day of use, Willison spent 10.42 in tokens, with the majority devoted to Datasette Agent development work. He suggests the model expanded knowledge is a reasonably good proxy for model size, indicating Fable may represent one of the largest models released to date.

8

CodexReleaseopenai/codex

0.138.0

## New Features - The `/app` command can now hand off the current CLI thread into Codex Desktop on macOS and native Windows, and Windows workspace launches can open directly into Desktop instead of stopping at a manual prompt. (#25638, #26500) - Local image attachments and standalone image generations now expose their saved file paths to the model, which makes follow-up edits and file references more reliable. (#25944, #25947) - Reasoning effort selection is more flexible: the TUI adds fallback shortcuts for terminals that miss `Alt` bindings, and model-defined effort levels now flow through in the order advertised by the model. (#25623, #26444, #26446) - App-server integrations can now read account token usage, and Codex auth supports v2 personal access tokens in CLI and app-server flows. (#25344, #25731) - Plugin automation got richer structured output: add/remove and marketplace commands support `--json`, plugin list JSON includes marketplace source, and plugin detail data now exposes default prompts, remote MCP servers, and unavailable app templates. (#26631, #26417, #25887, #26453, #26317) ## Bug Fixes - Goal workflows are more predictable: multiline paste in `/goal edit` no…

Claude CodeReleaseanthropics/claude-agent-sdk-python

v0.2.94

### Internal/Other Changes - Updated bundled Claude CLI to version 2.1.169 --- **PyPI:** https://pypi.org/project/claude-agent-sdk/0.2.94/ ```bash pip install claude-agent-sdk==0.2.94 ```

Claude CodeReleaseanthropics/claude-code

v2.1.169

## What's changed - Added `--safe-mode` flag (and `CLAUDE_CODE_SAFE_MODE`) to start Claude Code with all customizations (CLAUDE.md, plugins, skills, hooks, MCP servers) disabled for troubleshooting - Added `/cd` command to move a session to a new working directory without breaking the prompt cache mid-session - Added a `disableBundledSkills` setting and `CLAUDE_CODE_DISABLE_BUNDLED_SKILLS` environment variable to hide bundled skills, workflows, and built-in slash commands from the model - Fixed Up/Down arrows jumping to command history past the wrapped rows of a long input line — they now move through each visual row first, and history recall enters at the near edge - Fixed enterprise managed MCP policies (`allowedMcpServers`/`deniedMcpServers`) not being enforced on reconnect, IDE-typed configs, `--mcp-config` servers during the first session after install, or before remote settings loaded; also fixed slow cold starts for orgs without remote settings - Fixed a ~30-50ms UI stall at the start of each turn for macOS users logged in with claude.ai credentials - Fixed `claude -p` being slow or appearing to hang on Windows while waiting for the slash-command/skill scan (regression in 2…

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.169

## What's changed - Added an experimental `usage_EXPERIMENTAL_MAY_CHANGE_DO_NOT_RELY_ON_THIS_API_YET()` method on `Query` returning structured session cost, plan rate-limit, and local usage-behaviors data - Added an `sse` option (`SSEOptions`) to `BrowserQueryOptions` as an alternative to `websocket`, for browser SDK consumers who prefer Server-Sent Events ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.169 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.169 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.169 # or bun add @anthropic-ai/claude-agent-sdk@0.3.169 ```

CommunityNewsChizaram Ken / LogRocket

AI dev tool power rankings June 2026: OpenCode overtakes Cursor, Claude Opus 4.7 leads models

LogRocket analysis of AI dev tool standings in June 2026. Top Models: 1. Claude Opus 4.7 — Maintains leadership with highest WebDev Arena score (1567 Elo) 2. GPT-5.5 — New entry; OpenAI first fully retrained model since GPT-4.5 3. Qwen 3.7 Max — Surprise debut at #4 on WebDev Arena, priced at half cost of competing models 4. Claude Opus 4.6 — Drops one spot due to Qwen emergence 5. Claude Sonnet 4.6 — Holds steady as the accessible alternative Top Development Tools: 1. OpenCode — New leader; 160K+ GitHub stars, model-agnostic with 75+ providers 2. Cursor — Drops from #1; remains leading full-IDE experience 3. Claude Code — Holds as quality-focused professional option 4. Windsurf — Descends as OpenCode and Cursor advance 5. Antigravity — Remains top free choice during preview phase Notable: Five new models entered the field in June alone. DeepSeek V4 Pro offers frontier-level performance at significantly reduced pricing. OpenCode introduces unique LSP integration feeding compiler diagnostics back to models.

CommunityNewsMaya Posch / Hackaday

Hackaday: Revisiting AI Coding Assistants — You're Holding It Wrong Edition

Maya Posch revisits her previous critical assessment of AI coding assistants after receiving feedback about potential oversights. She addresses three key criticisms: using the wrong frontend (GitHub Copilot), selecting an inferior model (Claude Haiku 4.5), and employing improper prompting techniques. Model Performance Analysis: Examining coding benchmarks via LiveBench.ai, GPT-5.2 Codex scores highest at 83.62% accuracy, while Haiku 4.5 achieves only 72.17%. Posch notes that even in this scenario the best model only managed a rather pitiful 83.62%, suggesting significant reliability gaps across all options. Prompt Engineering Requirements: Effective use demands extensive environmental prompting — establishing detailed context and instructions before requesting code. The author questions whether this setup overhead negates the tool convenience, arguing that by the time you have described it thoroughly you might as well write the code yourself. Despite investigating better models and techniques, Posch maintains her skepticism about AI coding assistants practical value, particularly for complex, domain-specific programming tasks beyond simple boilerplate generation.

12

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.168

## What's changed - Updated to parity with Claude Code v2.1.168 ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.168 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.168 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.168 # or bun add @anthropic-ai/claude-agent-sdk@0.3.168 ```

Claude CodeReleaseanthropics/claude-code

v2.1.168

## What's changed - Bug fixes and reliability improvements

Claude CodeReleaseanthropics/anthropic-sdk-typescript

aws-sdk: v0.4.1

## 0.4.1 (2026-06-06) Full Changelog: [aws-sdk-v0.4.0...aws-sdk-v0.4.1](https://github.com/anthropics/anthropic-sdk-typescript/compare/aws-sdk-v0.4.0...aws-sdk-v0.4.1) ### Bug Fixes * **client:** run middleware before request signing ([#45](https://github.com/anthropics/anthropic-sdk-typescript/issues/45)) ([95f1a4a](https://github.com/anthropics/anthropic-sdk-typescript/commit/95f1a4a406c95aa31855fa1178fedfcb19978e16))

Claude CodeReleaseanthropics/anthropic-sdk-typescript

bedrock-sdk: v0.30.1

## 0.30.1 (2026-06-06) Full Changelog: [bedrock-sdk-v0.30.0...bedrock-sdk-v0.30.1](https://github.com/anthropics/anthropic-sdk-typescript/compare/bedrock-sdk-v0.30.0...bedrock-sdk-v0.30.1) ### Bug Fixes * **client:** run middleware before request signing ([#45](https://github.com/anthropics/anthropic-sdk-typescript/issues/45)) ([95f1a4a](https://github.com/anthropics/anthropic-sdk-typescript/commit/95f1a4a406c95aa31855fa1178fedfcb19978e16))

Claude CodeReleaseanthropics/anthropic-sdk-typescript

sdk: v0.102.0

## 0.102.0 (2026-06-06) Full Changelog: [sdk-v0.101.0...sdk-v0.102.0](https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.101.0...sdk-v0.102.0) ### Features * **api:** small updates to Managed Agents types ([8ba4f92](https://github.com/anthropics/anthropic-sdk-typescript/commit/8ba4f92df32eb434503b8c69bc7328adc5edcbec)) ### Bug Fixes * **client:** run middleware before request signing ([#45](https://github.com/anthropics/anthropic-sdk-typescript/issues/45)) ([95f1a4a](https://github.com/anthropics/anthropic-sdk-typescript/commit/95f1a4a406c95aa31855fa1178fedfcb19978e16))

Claude CodeReleaseanthropics/claude-agent-sdk-python

v0.2.93

### Internal/Other Changes - Updated bundled Claude CLI to version 2.1.167 --- **PyPI:** https://pypi.org/project/claude-agent-sdk/0.2.93/ ```bash pip install claude-agent-sdk==0.2.93 ```

Claude CodeReleaseanthropics/claude-code

v2.1.167

## What's changed - Bug fixes and reliability improvements

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.167

## What's changed - Updated to parity with Claude Code v2.1.167 ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.167 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.167 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.167 # or bun add @anthropic-ai/claude-agent-sdk@0.3.167 ```

Claude CodeReleaseanthropics/claude-agent-sdk-python

v0.2.92

### Internal/Other Changes - Updated bundled Claude CLI to version 2.1.166 --- **PyPI:** https://pypi.org/project/claude-agent-sdk/0.2.92/ ```bash pip install claude-agent-sdk==0.2.92 ```

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.166

## What's changed - Fixed MCP resource tools not being injected for servers added at runtime via the `mcp_set_servers` control request ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.166 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.166 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.166 # or bun add @anthropic-ai/claude-agent-sdk@0.3.166 ```

Claude CodeReleaseanthropics/claude-code

v2.1.166

## What's changed - Added `fallbackModel` setting to configure up to three fallback models tried in order when the primary model is overloaded or unavailable; `--fallback-model` now also applies to interactive sessions - Added glob pattern support in deny rule tool-name position (`"*"` denies all tools); allow rules reject non-MCP globs, and unknown tool names in deny rules warn at startup - Hardened cross-session messaging: messages relayed via `SendMessage` from other Claude sessions no longer carry user authority — receivers refuse relayed permission requests, and auto mode blocks them - `MAX_THINKING_TOKENS=0`, `--thinking disabled`, and the per-model thinking toggle now disable thinking on models that think by default via the Claude API (3P providers unchanged) - Claude Code now retries a turn once on the fallback model when the API rejects an unexpected non-retryable error; auth, rate-limit, request-size, and transport errors still surface immediately - `claude update` now announces the target version before downloading instead of going silent - `claude agents`: typing a URL into the list now filters to the session whose first prompt contained it - Fixed a recurring "image c…

12

Claude CodeReleaseanthropics/anthropic-sdk-typescript

aws-sdk: v0.4.0

## 0.4.0 (2026-06-05) Full Changelog: [aws-sdk-v0.3.1...aws-sdk-v0.4.0](https://github.com/anthropics/anthropic-sdk-typescript/compare/aws-sdk-v0.3.1...aws-sdk-v0.4.0) ### Features * **api:** mark Claude Opus 4.1 as deprecated ([4ee46f2](https://github.com/anthropics/anthropic-sdk-typescript/commit/4ee46f24cb2a5388027d7efb0b44fcb210ae8b50)) * **client:** add support for middleware ([9b01120](https://github.com/anthropics/anthropic-sdk-typescript/commit/9b011207965540211cefb19d44d758b7942aedab))

Claude CodeReleaseanthropics/anthropic-sdk-typescript

foundry-sdk: v0.3.0

## 0.3.0 (2026-06-05) Full Changelog: [foundry-sdk-v0.2.4...foundry-sdk-v0.3.0](https://github.com/anthropics/anthropic-sdk-typescript/compare/foundry-sdk-v0.2.4...foundry-sdk-v0.3.0) ### Features * **api:** mark Claude Opus 4.1 as deprecated ([4ee46f2](https://github.com/anthropics/anthropic-sdk-typescript/commit/4ee46f24cb2a5388027d7efb0b44fcb210ae8b50)) * **client:** add support for middleware ([9b01120](https://github.com/anthropics/anthropic-sdk-typescript/commit/9b011207965540211cefb19d44d758b7942aedab))

Claude CodeReleaseanthropics/anthropic-sdk-typescript

bedrock-sdk: v0.30.0

## 0.30.0 (2026-06-05) Full Changelog: [bedrock-sdk-v0.29.2...bedrock-sdk-v0.30.0](https://github.com/anthropics/anthropic-sdk-typescript/compare/bedrock-sdk-v0.29.2...bedrock-sdk-v0.30.0) ### Features * **api:** mark Claude Opus 4.1 as deprecated ([4ee46f2](https://github.com/anthropics/anthropic-sdk-typescript/commit/4ee46f24cb2a5388027d7efb0b44fcb210ae8b50)) * **client:** add support for middleware ([9b01120](https://github.com/anthropics/anthropic-sdk-typescript/commit/9b011207965540211cefb19d44d758b7942aedab))

Claude CodeReleaseanthropics/anthropic-sdk-typescript

vertex-sdk: v0.17.0

## 0.17.0 (2026-06-05) Full Changelog: [vertex-sdk-v0.16.1...vertex-sdk-v0.17.0](https://github.com/anthropics/anthropic-sdk-typescript/compare/vertex-sdk-v0.16.1...vertex-sdk-v0.17.0) ### Features * **api:** mark Claude Opus 4.1 as deprecated ([4ee46f2](https://github.com/anthropics/anthropic-sdk-typescript/commit/4ee46f24cb2a5388027d7efb0b44fcb210ae8b50)) * **client:** add support for middleware ([9b01120](https://github.com/anthropics/anthropic-sdk-typescript/commit/9b011207965540211cefb19d44d758b7942aedab))

Claude CodeReleaseanthropics/anthropic-sdk-typescript

sdk: v0.101.0

## 0.101.0 (2026-06-05) Full Changelog: [sdk-v0.100.1...sdk-v0.101.0](https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.100.1...sdk-v0.101.0) ### Features * **client:** add support for middleware ([9b01120](https://github.com/anthropics/anthropic-sdk-typescript/commit/9b011207965540211cefb19d44d758b7942aedab)) ### Bug Fixes * apply request timeout to inner fetch only, not middleware chain ([#40](https://github.com/anthropics/anthropic-sdk-typescript/issues/40)) ([25c13f6](https://github.com/anthropics/anthropic-sdk-typescript/commit/25c13f6183391f006396f16a9bd9a2c2f710d52c)) * **streaming:** carry stop_details through beta message_delta accumulation ([ed3fec7](https://github.com/anthropics/anthropic-sdk-typescript/commit/ed3fec7ace406fe4b0ca3494439061695dabf475)) * **streaming:** correctly parse json numbers with scientific notation ([#9](https://github.com/anthropics/anthropic-sdk-typescript/issues/9)) ([7d5e642](https://github.com/anthropics/anthropic-sdk-typescript/commit/7d5e6422dc5b58bc991c1587c4f1a0e8e753f2a5)) ### Chores * **internal:** fix artifact url ([925ec27](https://github.com/anthropics/anthropic-sdk-typescript/commit/925ec276fa97baf63d3643f24430b…

Claude CodeReleaseanthropics/claude-agent-sdk-python

v0.2.91

### Internal/Other Changes - Switched test suite from `pytest-asyncio` to anyio's pytest plugin, running every async test under both asyncio and trio backends to catch backend-specific regressions in CI (#1021) - Updated bundled Claude CLI to version 2.1.165 --- **PyPI:** https://pypi.org/project/claude-agent-sdk/0.2.91/ ```bash pip install claude-agent-sdk==0.2.91 ```

Claude CodeReleaseanthropics/claude-code

v2.1.165

## What's changed - Bug fixes and reliability improvements

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.165

## What's changed - Updated to parity with Claude Code v2.1.165 ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.165 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.165 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.165 # or bun add @anthropic-ai/claude-agent-sdk@0.3.165 ```

CursorChangelog

Design Mode Improvements

With Design Mode in the Cursor browser, you can click, draw, or describe changes by voice to help agents update your UI.

11

Claude CodeReleaseanthropics/claude-agent-sdk-python

v0.2.90

### Internal/Other Changes - Updated bundled Claude CLI to version 2.1.163 --- **PyPI:** https://pypi.org/project/claude-agent-sdk/0.2.90/ ```bash pip install claude-agent-sdk==0.2.90 ```

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.163

## What's changed - `stop_task` control requests now return success when the target task is already gone (`not_found` or `not_running`), so SDK clients can reliably prune stale task chips - Fixed SDK hosts being unable to add builtin MCP servers (e.g. `claude-in-chrome`) via `setMcpServers` when the CLI was launched without them - Stop and SubagentStop hook events now support `additionalContext` in `hookSpecificOutput`, enabling non-error feedback that continues the turn ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.163 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.163 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.163 # or bun add @anthropic-ai/claude-agent-sdk@0.3.163 ```

Claude CodeReleaseanthropics/claude-code

v2.1.163

## What's changed - Added `requiredMinimumVersion` and `requiredMaximumVersion` managed settings — Claude Code refuses to start if its version is outside the allowed range and directs the user to an approved version - Added `/plugin list` command to list installed plugins, with `--enabled`/`--disabled` filters - Added a "c to copy" shortcut to `/btw` that copies the raw markdown answer to the clipboard, preserving formatting when pasted elsewhere - Hooks: Stop and SubagentStop hooks can now return `hookSpecificOutput.additionalContext` to give Claude feedback and keep the turn going without being labeled a hook error - Skills: added `\$` escape syntax to include a literal `$` before a digit in command bodies - stdio MCP servers now receive the same `CLAUDE_CODE_SESSION_ID` as hooks/Bash on `--resume` - Fixed `claude -p` hanging forever after its final result when a backgrounded command never exits — background shells are now stopped ~5s after the result once stdin closes - Fixed `claude -p` failing with "ANTHROPIC_API_KEY required" on Bedrock/Vertex/Foundry when `CI=true` and no Anthropic API key is set - Fixed bash commands failing under bazel and EDR-protected Go workflows: `$TM…

CodexReleaseopenai/codex

0.137.0

## New Features - TUI controls now support F13-F24 keybindings, paste in searchable menus, and a compact reasoning-only status/title item (#25329, #25400, #25504). - Enterprise/admin flows now show monthly credit limits and can apply cloud-managed config bundles, including EDU workspaces (#24812, #24617, #24619, #24620, #24622, #25963). - Remote-control clients can start pairing and list or revoke controller grants through app-server v2 RPCs (#25675, #25785). - Plugin workflows gained machine-readable `codex plugin list --json` output and cached remote catalog suggestions (#25330, #25457). - Hosted web and image tools are available in more code-mode flows, with standalone web searches able to run in parallel (#25176, #25702, #25890, #25923). - Multi-agent v2 keeps runtime choice with each thread and exposes cleaner follow-up and metadata defaults for spawned agents (#25266, #25636, #25720, #25721, #25722, #25841, #26114). ## Bug Fixes - Cancelling a submitted prompt before visible output now restores the draft, attachments, and collaboration mode for editing (#25316). - Slash-command filtering and footer shortcut hints now reset or render according to the current UI state (#25492,…

CursorChangelogCursor

Cursor SDK — custom stores, custom tools, auto-review, nested subagents

TypeScript and Python SDK enhancements including custom tools, auto-review, and nested subagents. The Cursor SDK now supports custom stores for persistent agent state, custom tool definitions, automatic code review triggers, and subagents that can spawn further nested subagents.

6

Claude CodeReleaseanthropics/claude-agent-sdk-python

v0.2.89

### Internal/Other Changes - Updated bundled Claude CLI to version 2.1.162 --- **PyPI:** https://pypi.org/project/claude-agent-sdk/0.2.89/ ```bash pip install claude-agent-sdk==0.2.89 ```

Claude CodeReleaseanthropics/claude-code

v2.1.162

## What's changed - `claude agents --json` now includes `waitingFor` showing what a waiting session is blocked on (e.g. permission prompt) - `--tools`: explicitly listing Grep/Glob now provides the dedicated search tools on native builds with embedded search (previously these names were silently ignored) - `/effort` now confirms when your chosen level will persist as the default for new sessions - Clicking a slash command in the autocomplete menu now fills it into your prompt instead of running it immediately; press Enter to run - Remote Control now shows as a persistent footer pill (with a link to the session) instead of a startup message - Renamed Windsurf to Devin Desktop in the `/ide` menu, `/terminal-setup`, and `/scroll-speed`, following the editor's rebrand - Fixed a silent startup hang when the config directory is read-only or unwritable — Claude Code now starts with in-memory config and surfaces startup errors instead of showing a blank screen - Fixed WebFetch permission rules not being applied to built-in preapproved domains; explicit `WebFetch(domain:...)` deny/ask/allow rules now take precedence over the preapproved-host auto-allow - Fixed Windows permission rules neve…

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.162

## What's changed - Refusal error messages now carry `stop_reason: "refusal"` and `stop_details` on the assistant message and in session transcripts, so SDK consumers can detect refusals without text-matching the error content - Agent SDK sessions on native builds now default to fast embedded `find`/`grep` search in Bash, matching the interactive CLI, instead of always registering the dedicated Grep/Glob tools. To keep the dedicated tools (e.g. to intercept searches via `canUseTool` or hooks), name them in the `tools` option or reference them in `allowedTools` ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.162 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.162 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.162 # or bun add @anthropic-ai/claude-agent-sdk@0.3.162 ```

CursorChangelog

Organizations for Cursor Enterprise

Enterprise customers can now manage multiple Cursor teams from one place, with different security, governance, budget, and feature controls for each. These capabilities are now generally available to all Enterprise customers.

5

Claude CodeReleaseanthropics/claude-agent-sdk-python

v0.2.88

### Bug Fixes - **Trio compatibility for session stores**: Ported `session_store` code paths (`TranscriptMirrorBatcher`, `session_resume`, `sessions`) from raw `asyncio` primitives to `anyio`, fixing a crash (`TypeError: trio.run received unrecognized yield message`) when passing `session_store=` to `query()` or `ClaudeSDKClient` under trio (#990) ### Internal/Other Changes - Switched e2e CI jobs (`test-e2e`, `test-e2e-docker`, `test-examples`) from static API key to workload identity federation, using short-lived OIDC tokens with automatic refresh (#1018) - Updated bundled Claude CLI to version 2.1.161 --- **PyPI:** https://pypi.org/project/claude-agent-sdk/0.2.88/ ```bash pip install claude-agent-sdk==0.2.88 ```

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.161

## What's changed - The `initialize` control request is now idempotent: a second `initialize` returns the same success payload instead of an `Already initialized` error. `ControlResponse` gains an optional `pending_permission_requests` field, mirroring `ControlErrorResponse` - `applyFlagSettings` now live-applies `agent` changes: switching the active agent (or passing `null` to reset) takes effect on the next turn in a running session ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.161 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.161 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.161 # or bun add @anthropic-ai/claude-agent-sdk@0.3.161 ```

Claude CodeReleaseanthropics/claude-code

v2.1.161

## What's changed - `OTEL_RESOURCE_ATTRIBUTES` values are now included as labels on metric datapoints, so you can slice usage metrics by custom dimensions like team or repo - `claude agents` rows now show `done/total` before the detail when work is fanned out; peek shows the longest-running item - `/mcp` now collapses claude.ai connectors you've never signed in to behind a "Show unused connectors" row - Parallel tool calls: a failed Bash command no longer cancels other calls in the same batch — each tool returns its own result independently - Fullscreen mode: clipboard now uses `wl-copy`/`xclip`/`xsel` on Linux when available, copies to both the clipboard and PRIMARY selection for middle-click paste, and the "hold {key} for native selection" hint now shows the correct key per terminal - Fixed the `/effort` dialog, workflow animations, and prompt keyword shimmer not honoring the "Reduce motion" setting - Fixed `forceLoginOrgUUID`/`forceLoginMethod` managed-settings policies blocking third-party provider sessions (Bedrock, Vertex, Foundry, Mantle) alongside the org pin (regression in 2.1.146) - Fixed background subagent output corrupting `claude -p` stdout when using `--output-forma…

Claude CodeReleaseanthropics/claude-code

v2.1.160

## What's changed - Added a prompt before writing to shell startup files (`.zshenv`, `.zlogin`, `.bash_login`) and `~/.config/git/`, which could otherwise lead to unintended command execution - `acceptEdits` mode now prompts before writing build-tool config files that grant code execution (`.npmrc`, `.yarnrc*`, `bunfig.toml`, `.bazelrc`, `.pre-commit-config.yaml`, `.devcontainer/`, etc.) - Edit no longer requires a separate Read after viewing a file with `grep`: single-file `grep`/`egrep`/`fgrep` commands now satisfy the read-before-edit check - Fixed copy-on-select not writing to the Windows clipboard on WSL — now uses PowerShell interop instead of OSC 52, which terminals like MobaXterm don't support - Fixed restoring a completed session from `claude agents` dropping chat history and re-running the original prompt - Fixed background sessions re-attached after overnight retire losing their conversation and re-running the original prompt - Fixed `claude --bg` occasionally failing with "socket missing" when the background daemon was cold-starting on a loaded machine - Fixed an issue on Windows where the directory a background session was started in could not be deleted after `claude…

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.160

## What's changed - Fixed SDK hook callbacks swallowing abort signals: aborting during a PostToolUse hook now ends the turn with a final `result` message instead of hanging the calling process ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.160 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.160 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.160 # or bun add @anthropic-ai/claude-agent-sdk@0.3.160 ```

1

CodexReleaseopenai/codex

0.136.0

## New Features - TUI markdown now keeps web links clickable with OSC 8 metadata, and cramped tables switch to readable key/value records without losing link targets. (#24472, #24636, #24825) - Sessions can now be archived from the TUI with `/archive` or from the CLI with `codex archive` / `codex unarchive`; archived sessions are protected from resume/fork until restored. (#25027, #25021) - App-server integrations can resume a thread with its initial turns page, see richer MCP server status, and launch stdio mode with `codex app-server --stdio`. (#23534, #24698, #24940) - Remote execution setup now supports `CODEX_API_KEY` registration for approved OpenAI hosts, while remote-control websockets use short-lived server tokens instead of ChatGPT access tokens. (#24666, #24141) - Windows admins get an alpha `codex sandbox setup --elevated` provisioning path, plus requirements support for allowed Windows sandbox implementations. (#24831, #23766) - A feature-gated standalone image generation extension can run through the native Codex image artifact completion pipeline. (#24723, #24972) ## Bug Fixes - ChatGPT auth refreshes tokens before the five-minute expiry window and shows a relogin-r…

3

Claude CodeReleaseanthropics/claude-code

v2.1.159

## What's changed - Internal infrastructure improvements (no user-facing changes)

Claude CodeReleaseanthropics/claude-agent-sdk-typescript

v0.3.159

## What's changed - Updated to parity with Claude Code v2.1.159 ## Update ```sh npm install @anthropic-ai/claude-agent-sdk@0.3.159 # or yarn add @anthropic-ai/claude-agent-sdk@0.3.159 # or pnpm add @anthropic-ai/claude-agent-sdk@0.3.159 # or bun add @anthropic-ai/claude-agent-sdk@0.3.159 ```

This week for builders

Five minutes, every Monday — the tools, releases and tactics for shipping solo.

unsubscribe anytime.