Claude Blog 采集 (2026-05-31)¶
共采集 5 篇文章
📋 文章索引¶
- How to scale agentic coding across your engineering organization - Oct 15, 2025 (评分: 9.5)
- Generate better prompts in the developer console - May 20, 2024 (评分: 9.0)
- Improve your prompts in the developer console - Oct 14, 2024 (评分: 9.0)
- Claude takes research to new places - Apr 15, 2025 (评分: 9.0)
- Seeing like an agent: how we design tools in Claude Code - Apr 10, 2026 (评分: 9.0)
How to scale agentic coding across your engineering organization¶
来源: Claude Blog 发布日期: Oct 15, 2025 采集时间: 2026-05-31 价值评分: 9.5/10 正文字数: ~13360 字符
摘要¶
As Agentic coding tools mature, technical leaders are wrestling with a practical challenge: moving beyond isolated experiments to organization-wide adoption.
正文内容¶
As Agentic coding tools mature, technical leaders are wrestling with a practical challenge: moving beyond isolated experiments to organization-wide adoption.
The difference between successful and struggling implementations often comes down to execution. Teams that deploy agentic coding thoughtfully see meaningful improvements in development velocity and engineer satisfaction. Those that rush deployment without proper planning encounter resistance, inconsistent results, and difficulty demonstrating value. Working with engineering teams across different industries has surfaced common patterns. Successful adoption depends less on the specific tool and more on how you approach workflow changes, skill development, team dynamics, and success measurement. Let’s dive in. Understanding agentic coding capabilities Agentic coding tools differ from basic code completion by understanding broader context and handling multi-step tasks. They can plan approaches and work through implementation details with less hand-holding than earlier AI coding assistants. Common applications include: Legacy system modernization : Development teams use these tools to help migrate older codebases to current platforms. Projects that might have taken years can move faster, though they still require careful oversight and testing to preserve business logic correctly. Faster onboarding : New engineers can query codebases directly to understand architecture, dependencies, and implementation patterns. This complements traditional documentation and reduces the time before new hires contribute meaningfully. Incident response assistance : SRE and DevOps teams build agents that help diagnose and address common operational issues. While human oversight remains important for complex problems, routine incidents can often be handled with less manual intervention. Broader technical participation : Product managers can explore codebase constraints when writing requirements, and designers can create working prototypes from mockups. This doesn't replace engineering work but enables more informed collaboration across functions. These represent starting points rather than exhaustive possibilities for agentic coding applications. Planning your expansion approach Effective rollouts balance speed with learning. Rather than deploying to everyone at once or creating lengthy pilot phases, successful organizations build expertise incrementally while maintaining momentum. Start with super users Begin with a pilot group of 20-50 developers who already use AI-assisted tools. This group serves multiple purposes: validating the technology against your codebase, identifying useful workflows, and developing the internal expertise that will help broader adoption. Give your pilot group time to experiment with common use cases. Direct experience helps identify which customizations provide value and how well the tool integrates with your existing systems. Have them document patterns they discover—both what works and what doesn't. Practical pilot activities include: Creating custom slash commands for common tasks like database migrations or feature scaffolding Building CLAUDE.md files that capture coding standards and project-specific context Identifying repetitive workflows worth automating (boilerplate generation, test creation, dependency updates) Setting up a dedicated channel for troubleshooting and knowledge sharing Developing wrapper scripts for third-party tool authentication The pilot phase should surface both opportunities and challenges before you expand access more broadly. Launch with a hackathon Rather than a phased rollout where teams wait for access, consider uniting your organization with a kickoff event. Your pilot users can share techniques and prompts they've developed while everyone experiments together. This format helps demonstrate capabilities in a low-stakes environment. Engineers who are skeptical about AI assistance often change their perspective after hands-on experience. The collaborative atmosphere also surfaces creative applications your pilot group may not have considered. Keep the event accessible and energizing—food helps with both attendance and morale. Scale through internal expertise As more people use the tools, your pilot group transitions to an advisory role. They can run workshops, create educational content, and serve as resources when others encounter challenges. This approach tends to work better than external training programs because internal champions understand your specific environment and can provide relevant examples from actual projects. They speak your organization's language and know your particular pain points. Using CLAUDE.md files effectively CLAUDE.md files document repository conventions, environment setup, and project-specific behaviors. Their value grows when shared systematically across teams. Create project-level files : Check a CLAUDE.md file into your repository root. This ensures everyone working on the project inherits the same configuration and context automatically. Treat like documentation : Update CLAUDE.md files when architectural decisions change or new patterns emerge. Include these updates in pull requests alongside code changes. Include in onboarding : Make reviewing the project's CLAUDE.md file part of your developer onboarding checklist. New team members should understand both the codebase and how to use Claude Code within that context. Consider branch variations : For projects with significantly different patterns across branches, maintain branch-specific CLAUDE.md content that reflects each context. A typical project-level file might cover development environment requirements, testing and code standards, key architectural patterns, and current focus areas. This creates living documentation that keeps Claude Code aligned with your evolving practices. Measuring impact Pilots need clear success criteria. "How do we measure ROI?" remains a central question for driving adoption beyond early enthusiasts. Beyond lines of code written—which captures activity but not necessarily value—teams track multiple indicators: Sprint throughput : Teams with established DevOps practices can correlate adoption timing with changes in feature delivery speed. Task completion time : Measure how long standard tasks take before and after implementation. This granular view shows where agentic coding provides the most value. Migration velocity : Track time required to modernize legacy systems. Faster migrations free engineering resources for other priorities. Developer satisfaction : Survey engineers about time spent on repetitive versus creative work. Job satisfaction matters for retention and productivity. Onboarding duration : Measure how quickly new hires reach meaningful productivity. Shorter ramps reduce training costs and improve team capacity sooner. Cross-functional efficiency : Track how often other teams need dedicated engineering support for prototyping and testing. Reduced dependencies can indicate broader technical capability. Claude Code includes Activity Metrics that track lines of code accepted, suggestion acceptance rates, daily active users and sessions, organization-wide and per-user spending, and individual developer metrics. Sometimes the most persuasive measure is the simplest: concrete examples of tasks that now take a fraction of the previous time. When you can point to specific, meaningful efficiency gains, the value becomes self-evident. Common adoption challenges Several predictable issues emerge during agentic coding rollouts. Addressing them proactively improves outcomes: Scope tasks appropriately New users sometimes give agentic tools overly broad tasks without sufficient context, leading to frustrating results. Test-driven development provides helpful structure and clear success criteria. Start by writing tests that define what success looks like: required functionality, edge cases, error handling. Then implement features incrementally—just enough code to make one test pass at a time. For authentication, you might begin with basic login validation, then add password hashing, then session management. Run tests after each step and review the changes before proceeding. Claude Code can help analyze test results, but wait until current functionality works before expanding scope. Add new requirements gradually by writing tests first, then implementing to pass them. This prevents scope creep and maintains quality. Use focused commands like "write tests for user registration" followed by "implement the registration logic to pass these tests" rather than requesting everything at once. Provide adequate context Vague descriptions like "this isn't working" or "the button is too big" don't give the AI enough information to help effectively. Be specific: Share complete error information—full error messages, stack traces, and the specific action that triggered the issue. Copy terminal output or browser console errors directly into your session. Document your environment by including operating system, language versions, framework details, and relevant dependencies. The AI needs this context to provide accurate solutions. For UI issues, take screenshots and describe precisely what's wrong: "the login button extends 20 pixels beyond the container border on mobile screens" rather than "the button looks weird." Specify expected versus actual behavior clearly: "Expected: API returns 200 status with user data. Actual: Returns 401 with 'invalid token' message." Include relevant file contents—the specific code, configuration, or data related to your issue.
Develop effective prompting habits Communicating clearly with AI tools takes practice. Many developers expect immediate mind-reading and get frustrated when results miss the mark. Consider if a colleague would understand your request. If not, anticipate what questions they'd have and provide that information upfront. Structure requests with high-level goals first, then add implementation details. "Build a REST API for user management" followed by specific endpoints and requirements works better than mixing everything together. Use specific technical language instead of vague terms. "Optimize the database query to reduce response time from 2 seconds to under 500ms" beats "make it faster." Show what success looks like with concrete examples. "Follow this existing API pattern [paste code]" or "Use this coding style [share guide]" provides clearer direction than abstract requirements. Break complex work into sequential prompts: "Create the database schema," then "implement product catalog API," then "add shopping cart functionality." Each command should focus on one clear objective. Start simple and refine iteratively. "Create a basic user login form" followed by "add input validation" then "implement password strength requirements" tends to work better than specifying everything at once. Give specific feedback on output. "The error handling is too generic—add specific validation for email format and password length" guides improvement better than "fix the validation." Reference previous work explicitly when building on earlier steps: "Using the authentication middleware from earlier, now add role-based permissions." Moving forward Agentic coding shifts software development from writing every line to guiding implementation. Organizations that see good results focus on building foundations rather than rushing deployment. Start with a focused pilot group. Develop internal expertise. Build the infrastructure that supports success. Then expand deliberately through events like hackathons and internal champions. The path from pilot to production requires patience and systematic planning. Organizations that invest in this foundation tend to see meaningful returns: faster development, higher engineer satisfaction, and capacity to tackle previously difficult projects.
To start using Claude Code, you need to set up an API key and follow the documentation provided. This will guide you through the process of making requests and handling responses effectively.
Claude offers a range of features including natural language understanding, context awareness, and the ability to generate coherent and contextually relevant text. It can be used for various applications such as chatbots, content creation, and more.
Agentic coding tools like Claude Code can work for extended periods maintaining context and building on previous work. Rakuten's seven-hour autonomous refactoring session demonstrates sustained technical work without human intervention. The duration depends on your task complexity and project requirements.
Claude Code integrates with your existing development workflows through terminal integration. You can incorporate it into your current practices gradually, starting with specific tasks like testing or documentation before expanding to complete feature development.
Scale agentic coding across your engineering organization today.
Explore more product news and best practices for teams building with Claude.
Transform how your organization operates with Claude
Product updates, how-tos, community spotlights, and more. Delivered monthly to your inbox.
Please provide your email address if you'd like to receive our monthly developer newsletter. You can unsubscribe at any time.
采集自 Claude Blog，由 collect_claude_blog.py 自动采集
Generate better prompts in the developer console¶
来源: Claude Blog 发布日期: May 20, 2024 采集时间: 2026-05-31 价值评分: 9.0/10 正文字数: ~1255 字符
摘要¶
Claude now generates production-ready prompt templates, helping developers create effective prompts faster. You can now generate production-ready prompt templates in the Anthropic Console.
正文内容¶
Claude now generates production-ready prompt templates, helping developers create effective prompts faster.
You can now generate production-ready prompt templates in the Anthropic Console. Describe what you want to achieve, and Claude will use prompt engineering techniques such as chain-of-thought reasoning to create an effective, precise, and reliable prompt. This feature is designed to help users who are new to prompt engineering, as well as save time for experienced prompt engineers. You will get the best results by providing the prompt generator with detailed information about your task and desired output formatting. Although the generated prompts do not always produce perfect results, they often outperform hand-written prompts created by those who are new to prompt engineering. The generated prompt templates are also editable, allowing you to quickly tweak them for optimal performance.
Explore more product news and best practices for teams building with Claude.
Transform how your organization operates with Claude
Product updates, how-tos, community spotlights, and more. Delivered monthly to your inbox.
Please provide your email address if you'd like to receive our monthly developer newsletter. You can unsubscribe at any time.
采集自 Claude Blog，由 collect_claude_blog.py 自动采集
Improve your prompts in the developer console¶
来源: Claude Blog 发布日期: Oct 14, 2024 采集时间: 2026-05-31 价值评分: 9.0/10 正文字数: ~1249 字符
摘要¶
Claude can now automatically refine prompts in the developer console using techniques like chain-of-thought reasoning and example enrichment. Today, we're introducing the ability to improve prompts and manage examples directly in the Anthropic Console .
正文内容¶
Claude can now automatically refine prompts in the developer console using techniques like chain-of-thought reasoning and example enrichment.
Today, we're introducing the ability to improve prompts and manage examples directly in the Anthropic Console . These features make it easier to leverage prompt engineering best practices and build more reliable AI applications. Better prompts for better completions Prompt quality plays a significant role in how successful a model's responses are for a given task. However, prompting best practices take time to implement and often vary across model providers. The prompt improver allows developers to take existing prompts and leverage Claude to automatically refine them using advanced prompt engineering techniques. This is ideal for adapting prompts that were originally written for other AI models, as well as for optimizing hand-written prompts.
Explore more product news and best practices for teams building with Claude.
Transform how your organization operates with Claude
Product updates, how-tos, community spotlights, and more. Delivered monthly to your inbox.
Please provide your email address if you'd like to receive our monthly developer newsletter. You can unsubscribe at any time.
采集自 Claude Blog，由 collect_claude_blog.py 自动采集
Claude takes research to new places¶
来源: Claude Blog 发布日期: Apr 15, 2025 采集时间: 2026-05-31 价值评分: 9.0/10 正文字数: ~1175 字符
摘要¶
Claude can now conduct in-depth Research by searching across the web and your Google Workspace to deliver comprehensive answers in minutes. Earlier this year, we introduced our vision for Claude as your collaborative partner that delivers hours of work in minutes.
正文内容¶
Claude can now conduct in-depth Research by searching across the web and your Google Workspace to deliver comprehensive answers in minutes.
Earlier this year, we introduced our vision for Claude as your collaborative partner that delivers hours of work in minutes. In line with this vision, we’re continuing to expand the context that Claude has access to, as well as the kinds of outputs you can generate with Claude to support your personal and professional work. Today, we’re introducing two new capabilities that make Claude a more informed and capable collaborator — Research and a Google Workspace integration that connects your email, calendar, and documents to Claude. With Research, Claude can search across both your internal work context and the web to help you make decisions and take action faster than before.
Explore more product news and best practices for teams building with Claude.
Transform how your organization operates with Claude
Product updates, how-tos, community spotlights, and more. Delivered monthly to your inbox.
Please provide your email address if you'd like to receive our monthly developer newsletter. You can unsubscribe at any time.
采集自 Claude Blog，由 collect_claude_blog.py 自动采集
Seeing like an agent: how we design tools in Claude Code¶
来源: Claude Blog 发布日期: Apr 10, 2026 采集时间: 2026-05-31 价值评分: 9.0/10 正文字数: ~1991 字符
摘要¶
Building Claude Code: How Anthropic designs and refines AI agent tools like AskUserQuestion and Task tool. The key is progressive disclosure and learning to "see like an agent" to maximize effectiveness.
正文内容¶
Learn how the Claude Code team designs, tests, and evolves tools by thinking from the model's point of view.
One of the hardest parts about building an agent harness is constructing its tools. Claude acts completely through tool calling , but there are a number of ways tools can be constructed in the Claude API with primitives like bash , skills and code execution . (You can read more about programmatic tool calling on the Claude API in @RLanceMartin's new article ). So how do you design your agents' tools? Do you give it one general-purpose tool like bash or code execution? Or fifty specialized tools, one for each use case? To put yourself in the mind of the model, imagine being given a difficult math problem. What tools would you want in order to solve it? It would depend on your own skill set! Paper would be the minimum, but you’d be limited by manual calculations. A calculator would be better, but you would need to know how to operate the more advanced options. The fastest and most powerful option would be a computer, but you would have to know how to use it to write and execute code. This is a useful framework for designing your agent. You want to give it tools that are shaped to its own abilities. But how do you know what those abilities are? You pay attention, read its outputs, experiment. You learn to see like an agent. If you're building an agent, you'll face the same questions we did: when to add a tool, when to remove one, and how to tell the difference. Here's how we've answered them while building Claude Code, including where we got it wrong first. Improving elicitation with the AskUserQuestion tool
Explore more product news and best practices for teams building with Claude.
Transform how your organization operates with Claude
Product updates, how-tos, community spotlights, and more. Delivered monthly to your inbox.
Please provide your email address if you'd like to receive our monthly developer newsletter. You can unsubscribe at any time.
采集自 Claude Blog，由 collect_claude_blog.py 自动采集