Claude Blog 采集 (2026-08-18)¶
共采集 3 篇文章
📋 文章索引¶
- How ABC Legal turned every employee into a builder with Claude Managed Agents - Aug 17, 2026 (评分: 9.5)
- Securing the frontier: How JetBrains evaluates and deploys Claude Fable 5 - Aug 13, 2026 (评分: 9.5)
- Maximizing the value of your Claude Code sessions - Aug 14, 2026 (评分: 9.0)
How ABC Legal turned every employee into a builder with Claude Managed Agents¶
来源: Claude Blog 发布日期: Aug 17, 2026 采集时间: 2026-08-18 价值评分: 9.5/10 正文字数: ~7501 字符
摘要¶
ABC Legal transformed their organization’s AI adoption from scattered experiments to a governed fleet of specialized agents with Claude from Anthropic.
正文内容¶
ABC Legal transformed their organization’s AI adoption from scattered experiments to a governed fleet of specialized agents with Claude.
When Brandon Fuller, CTO of ABC Legal, a U.S.-based legal document delivery company, rolled out Claude Enterprise to the company's 1,100 employees earlier this year, something clicked immediately. Teams across the company (service of process, eFiling, and appearance counsel operations, plus marketing, compliance, finance, and more) started building automations on their own, without being asked. "Our users really flocked to it," Fuller recalls. "They saw the ease of use of connectors and tools, and suddenly we had people all over the organization automating the tasks that had always eaten up their day." It was exactly the kind of adoption any CTO hopes for. But Fuller saw an opportunity to go further: what if ABC Legal could also run a fleet of AI agents that were versioned, observable, and always on? That ambition came down to infrastructure. Early agents lived wherever their builder happened to put them, as scheduled tasks on individual desktops. Moving them off personal machines would let them run unattended and give Fuller a single view of what had been built, what it cost, and whether it ran last night. So he deployed Claude Managed Agents : one common deployment structure, shared workspaces, a single audit and billing surface, and always-on agents in the cloud instead of on a person’s laptop. As of July 2026, Fuller and his team at ABC Legal have tracked: 50+ agents built with Managed Agents in production Up to ~50% reduction in the cost of the human tasks some agents cover, before heavy optimization ~310 employees across every department using Claude for daily work Here’s how they got there and what they learned in the process. From enthusiasm to engineering: treating every agent like software When they first deployed Claude Managed Agents, Fuller had the team define every agent as code. He believes this is the natural form for an agent to take. As he explains, “an agent is really just structured text, a prompt plus configuration, and anything that is text can live in a repository where the whole company can see it, review it, and improve it.” An agent's prompt, tool list, schedule, credentials, and memory all go into configuration files kept in a git repository alongside the company's software. Nothing about an agent changes except through a pull request someone approves, which gives every agent version history, code review, rollback, and an audit trail. He spent a week building a starter kit with two templates, stored in dedicated git repositories. One is for event-driven agents, which start the moment something happens, like a new job arriving or a document coming back from a court. The other is for scheduled agents, which run on a timer: hourly, daily, or weekly. Each agent lives in its own folder with a standard structure: a JSON config file, a system prompt in Markdown, deployment scripts, and operational documentation. Merging a change into the main branch deploys the agent automatically. A builder never has to write software. They clone the repo, copy a starter template, tell Claude Code what the agent should do, and get back everything the agent needs: config, prompt, credential store, and memory. Bridging the technical divide Fuller gathered the company’s 15-person steering committee, drawn from finance, marketing, operations, and development (none of them software developers), and had them clone the repository and build Managed Agents using Claude Code. The goal was to prove that non-developers could build production agents themselves. If every agent had to route through the dev team, that bottleneck would cap how fast the whole company could move. What made it safe is that they were not writing software. Instead, they were filling in configuration and a prompt, and Managed Agents supplied the runtime. "I had to explain what a PR was to them. A lot of [the non-software engineers] thought it meant running, like a PR, the fastest you can,” he said. “Now they're doing pull requests and sending them to each other." Within a week, all 15 employees had working agents. Those builders went back to their teams and trained others. Within a month, roughly 50+ agents were running across ABC Legal. Each agent has a name, an owner, and a single job. An agent for most stages of the legal document process ABC Legal now has an agent at most stages of the legal filing process and the operations around it. The AI Code Reviewer reviews every pull request across four codebases, running multi-model analysis to catch security bugs, performance regressions, and committed credentials. Engineers now wait for its review before merging. The EvidenceChain™ Delivery Agent took over a weekly chore an account manager used to do by hand. ABC Legal runs a proprietary site, EvidenceChain.com, where courts, plaintiffs, and defendants look up the record of a service completed in the field, including who the process server was, when they attempted it, and photos of the document delivery. One customer wanted specific records pulled from it on an ongoing basis. The agent now pulls a database report for matching jobs, retrieves each PDF with a browser built into the Managed Agent, and delivers it to the customer's FTP server daily. The account manager who set it up had never automated anything, and built it in about an hour by describing it to Claude Code. The eFiling Rejection Diagnoser fires automatically when a court rejects a filing, reads the job details, checks the court's rules, and posts a diagnosis to Slack in about a minute, work that used to consume hours of an employee’s day. A job-verification agent checks every incoming job against the courts. It navigates a court website in a browser, confirms the hearing or case is filed appropriately and actually occurring on the stated date, then adjusts the job based on what it found, flagging jurisdictions, courts, and statute-of-limitations timeframes. The Attorney Coverage Agent works the network of attorneys to get hearings covered, checking availability, emailing them, and reading replies about availability and pricing so a coordinator can confirm coverage. In finance, an AR-remittance agent parses a remittance email, builds the NetSuite payment-application file, and posts it to Slack for one-click approval, and then imports it, with a daily agent that renders a capitalize-or-expense verdict on each engineering ticket. Marketing runs a Google Ads analyst that posts a weekly recommendation for the channel lead. In operations, a review agent called Charvis checks completed service jobs and now agrees with the compliance team about 98% of the time. The Service-Overdue-Nudger works the tier-1 layer of ABC Legal's operational backlogs, the repetitive first pass a person would otherwise do, and drafts tiered daily outreach messages for human approval. Making the agents smarter: harvest, tune, repeat ABC Legal's agents work under human supervision, posting what they did or what they recommend to Slack, where people reply in threads and react with emoji.
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
Securing the frontier: How JetBrains evaluates and deploys Claude Fable 5¶
来源: Claude Blog 发布日期: Aug 13, 2026 采集时间: 2026-08-18 价值评分: 9.5/10 正文字数: ~6287 字符
摘要¶
How JetBrains uses Claude Fable 5 and why they view safeguards and data retention as central to working with frontier models.
正文内容¶
JetBrains Agent Systems CTO Vladislav Tankov on how the company evaluates frontier models against its private repositories, when his team reaches for Claude Fable 5, and why he treats safeguards and data retention as central to working with them.
JetBrains builds the tools developers use worldwide, from IntelliJ IDEA and PyCharm to the Kotlin programming language, serving more than 12.5 million active users and 88 of the Fortune Global 100. Vladislav Tankov, CTO at JetBrains, spoke with Anthropic about how his team evaluates new models, decides when to use Claude Fable 5, and thinks about data retention and safeguards when working with frontier models. How has frontier AI changed for JetBrains in 2026? I've been with JetBrains for 10 years, and we were among the very first customers of LLM providers. Over the last year, we moved from having AI skeptics among our customers and inside the company to seeing that AI is here to stay. It's a big and foundational change in the technology industry. Literally every skeptic in the company has changed. How do you evaluate new models and decide when to use them? We're a coding company, so we have a big evaluation pipeline: large eval sets on private repositories, including our monorepo. We take a close look at whether a model lives up to its benchmark scores on real work—some models are tuned to score well on public benchmarks but fall down on actual tasks. With a private repository, that's a lot easier to check. We also keep leaderboards for best quality, best cost per task, and fastest model. While Claude Fable 5 is more expensive per token, its cost per task is lower in some cases, particularly for more complicated, long-running work. How did Claude Fable 5 score on your evals relative to previous models? Claude Fable 5 is both more accurate and more efficient than prior models. It posted the best Python pass rate in our suite at 44.3%, against 28.2% for Opus 4.8, a 16-point jump. In a head-to-head comparison, Claude Fable 5 solved 18 Python tasks that Opus 4.8 missed and lost only 2. Its answers are also more trustworthy: when its code ran, it passed our tests far more often than either Opus model. That matters because code that runs but produces wrong answers is the most expensive kind of failure to catch. The efficiency story is just as interesting. Claude Fable 5 needed about 22% fewer steps than Opus 4.8 to reach a solution, so it gets to working code with less trial and error. It also spends its effort in the right places. On Java tasks, Opus 4.8 repeatedly tried to pull in outside resources that almost never help in our environment, while Claude Fable 5 skipped that entirely and worked with the code in front of it. It shows better engineering habits more generally. When do you use Claude Fable 5 over other models? Opus is seen as a workhorse: you can be very sure it will do the work. You go to Claude Fable 5 when you really need good reasoning, when you almost need a partner, and you're not sure yourself how to do the thing. For example, one of our tech leads decided to implement a rich text editor component we had attempted a few times over the years, and Claude Fable 5 almost one-shotted it. Another popular Claude Fable 5 use case is long-running agentic-coding experimentation. We provide an agent running Claude Fable 5 with specifications (in the form of text and images) and make it implement sophisticated IDE-like apps. The interesting thing here is that specifications can also be generated by the agent, based on the existing app. Joining these two components allows us to rewrite the app from one runtime, framework, or language to another in a nearly black-box setup. How are you thinking about safety and data retention with today's frontier models? We're not a company trying to create the safest model ourselves. We expect that the red teaming and everything else done on Anthropic's side is enough to believe the model is safe. Then we take a systematic approach to deployment, where we can guarantee safety: creating the infrastructure and the safety net around the model and the harness, rather than tweaking the model itself. Security is also one of our biggest Claude Fable 5 uses. We run white-box testing against our own products to find vulnerabilities, and our security team is preparing for the fact that not only are we running the model—people outside the company will be running Claude Fable 5, or similar-class models, to probe for vulnerabilities across all of our products. Since we serve large enterprises in regulated industries, it's important for us to be prepared. Claude Fable 5 supports our work rather than blocking. So it's a tight balance: the less aggressive the classifier is on your side, the more vulnerabilities someone will find in our products—including ones nobody knew about. And it's no secret: we'd prefer zero data retention. But I don't see any other way for you to understand what was asked and where a classifier may have worked incorrectly. As long as reviews are only to investigate the most serious cases flagged , I'm okay with it. I think it's a fair tradeoff for access to frontier intelligence that allows my team to do their best work. What's next on JetBrains's AI roadmap? We expect the underlying models built by the LLM providers to keep getting more capable. What matters now is a kind of cockpit for software development: a space in which agents and people collaborate, and where people can manage the development process. For JetBrains, it’s a big transformation. We see an opportunity to build the next generation of products across the agentic software development lifecycle that powers that cockpit. Developers will get more and better code shipped with agents, non-technical roles will have a larger role in software creation, and organisations will get the governance and clarity on the return on investment they need. Get started with Claude Fable .
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
Maximizing the value of your Claude Code sessions¶
来源: Claude Blog 发布日期: Aug 14, 2026 采集时间: 2026-08-18 价值评分: 9.0/10 正文字数: ~2722 字符
摘要¶
Practical tips for how to run efficient sessions that get the most value from every token.
正文内容¶
How to run efficient sessions that get the most value from every token.
TL;DR Run /clear between tasks. This prevents prior irrelevant context from being sent back to the model, which can reduce token usage. Set your model and effort level before you start. Changing either one mid-conversation can bust your prompt cache, which can increase token cost. @-mention files instead of naming them. The file gets attached to your message directly, which saves a Read call, or a search if Claude has to go find it. Add quiet flags to noisy commands, or run them in a subagent. Command output is added to the conversation just like a file, and stays there for the rest of the session. Run /context once in a fresh session. It shows what's loaded ( CLAUDE.md , MCP tool definitions), so you can cut out anything unnecessary. /compact before you take a break from your keyboard. The prompt cache expires after an hour, and summarizing a conversation is much cheaper while it's still cached. Maximizing value Until pretty recently, the tools you wrote code with were a flat fee (or free). Your editor cost the same whether you fixed one test or fifty that afternoon, so an individual task didn't really have a price of its own. With agentic coding tools like Claude Code, it does. The same completed task can also cost different amounts depending on how you use it. In one session, Claude reads the test and the file it covers, makes the edit, and is done in a handful of turns. In another, it greps around the repo first, reads a dozen files on its way to the same two, and every one of those turns also drags along everything else that's been read into the conversation since this morning.
Input and output tokens A request goes through the GPU in two phases, and they cost different amounts. First, during prefill, the model reads your request and context: the system prompt, your CLAUDE.md , your message, and everything that's been added to the conversation since (the files Claude has read and the output of the commands it ran). Those are your input tokens. Then, during decode, it writes output tokens: its thinking, the tool calls it makes, and the text you see. This happens one token at a time; a 200-token response is 200 runs of the model, one after the other. Per token, decode keeps the GPU busy for a lot longer, which is why output is priced at roughly 5x input.
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