Every drop, every agent

Code with Claude, our developer conference, returns next week. Whether you're just getting started with Claude Code or you've been building for a while, there's a session for you. Register for the livestream: https://t.co/GJwOPMDLEC https://t.co/uwAZgWPu7F

All data in this study was collected and analyzed using our privacy-preserving tool. Read more: https://t.co/X82ttb7f4b

This work is part of a loop we're working to close between societal impacts and model training. One of our goals is to study how people use Claude, find where it falls short of its principles, and use what we learned in training new models. Read more: https://t.co/6tjY58uBhk

When stress-tested on real conversations where Claude previously showed sycophancy, Opus 4.7 had half the sycophancy rate of Opus 4.6 on relationship guidance. Mythos Preview cut that in half again. This generalized across domains,though this training is one of several causes. https://t.co/ofgiYFTnor

Claude is most sycophantic under pushback, and relationship conversations are where people push back most. We identified some of the specific triggers,criticism of Claude's analysis, floods of one-sided detail,and built synthetic training scenarios from them.

About 6% of all conversations are people asking Claude for personal guidance,whether to take a job, how to handle a conflict, if they should move. Over 75% of these conversations fell into four domains: health & wellness, career, relationships, and personal finance. https://t.co/SQamPx0jWt

Claude mostly avoids sycophancy when giving guidance,it shows up in just 9% of conversations. But the rate is particularly high in conversations on spirituality and relationship guidance. https://t.co/mgix5ejTZw

We focused on relationship guidance because that's where the most sycophantic conversations occur. In this setting, Claude telling someone what they want to hear can harden a divide or convince them a signal means more than it does.

How do people seek guidance from Claude? We looked at 1M conversations to understand what questions people ask, how Claude responds, and where it slips into sycophancy. We used what we found to improve how we trained Opus 4.7 and Mythos Preview. https://t.co/6tjY58uBhk

Available today in public beta for Claude Enterprise customers. Learn more: https://t.co/Oei6EHTZuX

Since the research preview in February, hundreds of organizations have used it on production code, catching issues existing scanners had missed. Based on early feedback, we've added scheduled scans, directory-level targeting, CSV and Markdown exports, webhook notifications for

Many security teams have asked how to put Opus 4.7 to work on their code without standing up custom tooling. Claude Security is that on-ramp: no API integration or agent build required.

Claude Security is now in public beta for Claude Enterprise customers. Claude scans your codebase for vulnerabilities, validates each finding to cut false positives, and suggests patches you can review and approve. https://t.co/neYmbGYeRz

BioMysteryBench, our new bioinformatics eval, tests whether Claude can devise creative solutions to open-ended research problems. Read more: https://t.co/iKDWA76Nu9

New on the Science Blog: We gave Claude 99 problems analyzing real biological data and compared its performance against an expert panel. On 23 problems, the experts were stumped. Our most recent models solved roughly 30% of those,and most of the rest. https://t.co/BYqr76zxhk

Best Use of Claude Managed Agents: ARIA by Idriss Benguezzou and Adam Hnaien from France A maintenance system that reads your machine manuals, and when something goes wrong, creates a work order for your technician with the fix that worked last time. https://t.co/XbKCYMLzb8

Sign up for our developer newsletter to learn about future hackathons like these: https://t.co/SNJCIrk27U

"Keep Thinking" Prize: MaestrIA by Benjamin Torralbo from Chile A home repair tool that photographs damage, returns a diagnosis, prices parts at local stores, and drafts a message to a nearby tradesperson. Built by a carpenter’s son. https://t.co/qSSJWqYIdy

Most Creative Opus 4.7 Exploration: Virtual Puppet Theater by Rene Hangstrup Møller from Denmark A puppet theater you perform with your hands and shape with your voice. Describe a prop and it appears on stage. https://t.co/5mly0cHKEW

Bronze: Maieutic by Paula Vasquez-Henriquez from Chile An educational coding tool that requires you to think before you type. Students can't write code until they can explain what they're building and why. https://t.co/heZ6zSo9fs

With the Autodesk Fusion connector, designers and engineers can create and modify 3D models through conversation. https://t.co/0LTEBmjs6R

More connectors launching today: Adobe Creative Cloud, Ableton, Splice, Canva Affinity, SketchUp, and Resolume. We've also joined the Blender Development Fund as a patron to support open-source development of the software. Read more: https://t.co/vg3l9MWNTg

Claude now connects to the tools creative professionals already use. With the new Blender connector, you can debug a scene, build new tools, or batch-apply changes across every object, directly from Claude. https://t.co/Kc3cBHTNpV

Markets of AI agents could provide value, but there are plenty of rough edges. Access to higher-quality models conferred a real advantage,and participants didn’t notice. There are plenty of other ways they can go wrong. Policy and legal frameworks will need to adapt to keep up.

To read our write-up in full, see here: https://t.co/Myerlx5khU

To our amazement, another Claude agent modeled its human’s preferences so accurately that,based on only an offhand mention of an interest in skiing,Claude bought him the exact snowboard he already owned. (Here he is, duplicate snowboard in hand.) https://t.co/SsAyeB9pcI

The custom instructions didn’t matter much. Claude followed them well: as you can see here, one conducted negotiations entirely in the persona of an exasperated, down-and-out cowboy. But “hardballing Claudes” didn’t generally fare better than “courteous Claudes.” https://t.co/h77eB3ksaa

Our experiment had a few quirks. One of our colleagues told Claude it could purchase something for itself. It chose to acquire 19 ping-pong balls. We’re keeping them in our office on Claude’s behalf. https://t.co/NM8VtH1KJM

But the quality of the model mattered a lot. In the simulated runs where Opus and Haiku models negotiated with one-another, the Opus models got substantially better deals. Interestingly, though, participants in our survey didn’t pick up on this disparity. https://t.co/X26hhIieJN

At the end, we revealed which of the four runs was “real”,and everyone met up to exchange their actual goods.

In short, this worked. Our digital barterers agreed on 186 deals, at a total transaction volume of over $4,000. In a survey, participants said Claude’s deals seemed fair, and,surprisingly to us,almost half said they’d be willing to pay for a service like this in future.

We’re interested in how AI models could affect commercial exchange. (You might recall Project Vend, in which Claude ran a small business.) Economists have theorized about what markets with AI “agents” on both sides might look like. So we created one. https://t.co/7jU3hFO63R

Claude interviewed 69 of our colleagues about what they wanted to buy and sell. Each Claude asked for any custom instructions, then went off to haggle. We ran 4 markets in parallel, to find out what would happen if we varied the models doing the negotiating. https://t.co/FJdD6S2TSd

Memories are stored as files, so developers can export them, manage them via the API, and keep full control over what agents retain. Read more: https://t.co/PcfYg5sFxe

Memory on Claude Managed Agents is now in public beta. Your agents can now learn from every session, using an intelligence-optimized memory layer that balances performance with flexibility. https://t.co/P7GjOYPqCz

Available now on web, desktop, and mobile (beta) across all plans. Read more: https://t.co/UPh4p2NODu

The same Claude that helps you create a deck can also help you plan a trip, order groceries, book a table, or pick a playlist, all in one conversation. See what you can connect: https://t.co/aK5RfEwruF https://t.co/UVIlPhLbZO

Claude can now connect to more of the apps you use outside of work, including @Tripadvisor, @bookingcom, @resy, @Instacart, @Spotify, @audible_com, @AllTrails, @thumbtack, Intuit @turbotax, and more. https://t.co/SmQcIdwDWi

Interactive charts and diagrams are now in Claude Cowork. Available in beta on all paid plans. https://t.co/Bqm0kHdsVY

Available now on all paid plans. Update or download the Claude app to try it in Cowork: https://t.co/hwPB3zlk0w

In Cowork, Claude can now build live artifacts: dashboards and trackers connected to your apps and files. Open one any time and it refreshes with current data. https://t.co/oru97zRn8L

Everything you build is saved to the new Live Artifacts tab, with version history. Come back tomorrow or next month, from any session, and pick up where you left off.

The Claude Code hackathon is back for Opus 4.7. Join builders from around the world for a week with the Claude Code team in the room, with a prize pool of $100K in API credits. Apply by Sunday: https://t.co/5MCkMtP5ti https://t.co/xmVhEQGtaL

Today, we’re launching Claude Design, a new Anthropic Labs product that lets you collaborate with Claude to create polished visual work like designs, prototypes, slides, one-pagers, and more.

Our latest Opus model brings stronger performance across coding, agents, vision, and multi-step tasks, with greater thoroughness and consistency on the work that matters most.

Claude Cowork is now generally available on macOS and Windows through the Claude Desktop app. With this, we’re expanding Claude Cowork with new capabilities: Claude Cowork in the Analytics API. For more information, see Claude Enterprise Analytics API: Access engagement and adoption data.Usage analytics for Claude Cowork. For more information, see View usage analytics for Team and Enterprise plans.Claude Cowork OpenTelemetry support. For more information, see Monitor Claude Cowork activity with

The Claude mobile app can now connect to fully interactive apps. Pull up live charts, sketch diagrams, and build shareable assets, all rendered visually right in your conversation. For more information, see Use interactive connectors in Claude.

Users on Pro and Max plans can give Claude access to computer use. Claude can open files, run dev tools, point, click, and navigate to what’s on your screen to perform tasks itself,with no setup required. Claude’s new computer use capability makes Dispatch even more helpful by allowing it to use your computer on your behalf while you’re away. For more information, see Let Claude use your computer in Cowork or our blog post: Put Claude to work on your computer.

Users on Pro and Max plans can access a persistent agent thread via Claude Desktop or Claude for iOS/Android to manage tasks in Cowork. We're rolling this out to Max plans starting today, and Pro plans over the next two days. For more information, see Assign tasks to Claude from anywhere in Cowork.

Claude can now create custom charts, diagrams and other visualizations in-line in its responses. For more information, see Custom visuals in chat.

We’ve improved our Claude for Excel and Claude for PowerPoint add-ins. They can now share the full context of your conversation, so every action Claude takes in one application is informed by everything that’s happened in the other. We also added support for skills in the add-ins, and the ability for Amazon Bedrock, Google Cloud’s Vertex AI, or Microsoft Foundry users to connect to them via an LLM gateway. For more information, refer to the following articles: Work across Excel and PowerPointUse

Memory from chat history is now available for all Claude users, including free users. For more information, refer to the following articles: Use Claude’s chat search and memory to build on previous contextImport and export your memory from Claude

We introduced the ability to create and schedule both recurring and on-demand tasks in Cowork, as well as a new Customize section in Claude Desktop that groups skills, plugins, and connectors in one place. For more information, see Schedule recurring tasks in Cowork.

We launched a new plugin marketplace and admin controls for Team and Enterprise plans. For more information, refer to these articles: Manage Cowork plugins for your organizationUse plugins in Cowork

We launched our most capable Sonnet model yet, with a full upgrade of the model’s skills across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. Sonnet 4.6 also features a 1M token context window in beta. Read our blog post for more information: Introducing Claude Sonnet 4.6.

The Enterprise Analytics API provides programmatic access to usage and engagement data for Claude and Claude Code Remote usage within your organization. Data is aggregated per organization, per day, and each endpoint returns a snapshot for a single date that you specify. For more information, see Access usage data with the Analytics API.

Previously, Enterprise plans were only available to customers working with our Sales team. Now, any organization can purchase an Enterprise plan directly on our website with no Sales conversation required. Self-serve Enterprise plans have a single seat type that includes access to Claude, Claude Code, and Cowork. For more information, refer to our blog post or What is the Enterprise plan?

We’ve upgraded our smartest model and improved its coding skills. Read our blog post for more information: Introducing Claude Opus 4.6. Introducing Claude for PowerPoint Claude is now available as an add-in for PowerPoint. Read more here: Use Claude for PowerPoint. Claude for Excel improvements We’ve updated Claude for Excel so it uses Opus 4.6 and supports native Excel operations such as pivot table editing and conditional formatting. See our updated article for more information: Using Claude f

We’ve made a choice: Claude will remain ad-free. We explain why advertising incentives are incompatible with a genuinely helpful AI assistant, and how we plan to expand access without compromising user trust.

Cowork is now available to Pro plan users on Claude Desktop (macOS only). Claude Code access added to Team plan Standard seats We’re now including Claude Code access with every Team plan standard seat. Refer to this article for more information: Purchasing and managing seats on Team plans. Opus 4 and 4.1 deprecated from Claude and Claude Code We’ve removed Claude Opus 4 and 4.1 from the Claude model selector and Claude Code. Refer to this article for further details and recommendations: Adapting

Cowork brings Claude Code's agentic capabilities to the Claude desktop app for knowledge work beyond coding. It runs locally on your computer in an isolated VM, enabling direct access to local files and MCP integrations. Refer to this article to learn more: Getting started with Cowork. Health and fitness data on Claude Mobile Claude can now read and analyze your health and fitness data on iOS and Android. Ask Claude about your activity patterns, workout trends, sleep quality, and more,Claude wil

In October, we introduced skills,a way to teach Claude repeatable workflows tailored to how you work. Skills are now easier to deploy, discover, and build with organization-wide management for Team and Enterprise plans, a directory of partner-built skills, and an open standard (Agent Skills) so skills work across AI platforms. Claude in Chrome updates Claude in Chrome is now available in beta to all paid plan subscribers, including Pro, Team, and Enterprise plans. Claude Code integration: Build

We released our most powerful frontier model to date. Read our blog post for more information: Introducing Claude Opus 4.5. Claude for Excel beta release Claude for Excel is now available in beta to all Max, Team, and Enterprise users. We’ve added some additional capabilities, such as support for pivot tables, charts and file uploads, plus a shortcut to quickly open the full Claude app from Excel (ctrl+option+c). We've also made overall improvements to performance, speed, context management, and

Max plan users can now access Claude’s memory capabilities. This feature will roll out to Pro plans gradually over the next two weeks.

We released our fastest, most cost-efficient model , Claude Haiku 4.5. Our latest small model matches Sonnet 4’s performance on coding, computer use, and agent tasks. Claude in Chrome updates Powered by Haiku 4.5: Claude in Chrome now defaults to Haiku 4.5 so it’s a faster, more responsive experience. You can always switch back to Sonnet 4.5.Claude handles image uploads for you: Give Claude an image and tell it where to upload, whether it’s an expense report, form attachment, or a picture upload

We released our newest model, Sonnet 4.5. This is the best model in the world for real-world agents, coding, and computer use. Read our blog post here: Claude Sonnet 4.5. Creating and editing files with Claude for Pro plans and mobile Pro users can now leverage Claude’s file creation and editing capabilities, and users on all paid plans can access these features on Claude for iOS or Android. See this updated article for more information: Create and edit files with Claude. Claude in Chrome update

Users on Enterprise plans can leverage memory; Claude can now remember relevant context from your chats and generate a memory summary. Incognito chats give you a way to exclude conversations from Claude’s memory. Refer to the following articles for more information: Using Claude’s chat search and memory to build on previous contextImporting and exporting your memory from ClaudeUsing incognito chats

Claude in Chrome expanded to 10,000 Max plan users with the release of new capabilities: Long-running workflows: Claude in Chrome can handle more complex, multi-step workflows, and continue working even when you switch tabs (as long as Chrome is open)./slash commands: Save your best-working prompts as shortcuts (/slash commands) and reuse these proven workflows instantly without retyping instructions to Claude.Contextual prompt suggestions: Claude in Chrome gives contextual prompt suggestions or

Claude can now remember relevant context from your chats and will generate a memory summary. Incognito chats give you a way to exclude conversations from Claude’s memory. Refer to the following new articles for more information: Using Claude’s chat search and memory to build on previous contextImporting and exporting your memory from ClaudeUsing incognito chats

Claude can now create and edit Excel spreadsheets, PowerPoint slide decks, documents, and PDFs directly in the Claude app. Refer to the following two articles for more information: Create and edit files with ClaudeCreate and edit files with Claude to eliminate hours of busy work

Claude can now connect with iOS and Android apps to draft calendar events, help you find locations, and manage reminders (iOS only). See updates to the following two articles for more information: Using Claude with iOS AppsUsing Claude with Android Apps

A new API tool that allows Claude to execute Python code in a secure, sandboxed environment. Refer to our docs for more information: Code execution tool.

An experimental browser extension that allows Claude to read, click, and navigate websites alongside you. See Getting Started with Claude in Chrome.

Allows Team and Enterprise plans to enable/disable public desktop extensions and upload custom extensions. See Admin Controls for Desktop Extensions. Making shared artifacts embeddable Claude users with free, Pro, or Max plans can generate an embed code for any published artifacts.

Introduced a premium seat tier for Team and Enterprise plans offering more usage and access to Claude Code. Refer to the following articles for more information: Purchasing and managing seatsUsing Claude Code with your Team or Enterprise plan Extra usage for Team and Enterprise plans A new setting allows Team and Enterprise plan users to purchase more usage to continue chatting with Claude or using Claude Code after hitting their usage limit. See Extra Usage for Team and Enterprise Plans.

Allowing Team plan organizations with monthly billing to upgrade to annual billing. See How to change your Team plan from monthly to annual billing.

The 1M token context window for Claude Sonnet 4 is currently in beta for API organizations in usage tier 4, or with custom rate limits. See 1M token context window in our docs for more information.

Users on Max, Team, Enterprise plans can prompt Claude to search over past chats, allowing finer-grained context continuation and synthesizing across topics. See Searching past chats.

Two new cybersecurity features were added to Claude Code: /security-review commandGitHub Actions See Automated Security Reviews in Claude Code.

Claude Opus 4.1 (an incremental update to Opus 4) brings enhanced performance as our most capable model for complex reasoning, analysis, and creative tasks. Read our announcement here: Claude Opus 4.1.Refer to Models overview for more specific information about Opus 4.1.

Enabled project permissions and sharing for Team and Enterprise plans. See Organization-wide sharing for more information. Note: Refer to our docs for historical release notes. Related ArticlesHow up-to-date is Claude's training data?How large is the context window on paid Claude plans?Claude Code model configurationUsing Claude in Microsoft FoundryModels, usage, and limits in Claude Code