Anthropic 已将其面向 Claude Code 的自动化拉取请求审查工具 Code Review 从研究预览阶段推出，面向 Team 和 Enterprise 客户提供，符合条件的企业最高可获 5 万美元免费积分。采用数据驻留配置的 GitHub Enterprise Cloud 组织——即欧洲企业为将代码留在欧盟所采用的配置——不在优惠范围内。
Anthropic has moved Code Review, its automated pull-request reviewer for Claude Code, out of research preview for Team and Enterprise customers, with up to $50,000 in free credits for eligible enterprises. Organisations running GitHub Enterprise Cloud with data residency, the configuration European firms adopt to keep code in the EU, are excluded from the offer.
Anthropic 已将其自动化代码审查工具正式推出，向企业客户提供最高 5 万美元积分。将代码留在欧洲的组织无法领取任何积分。
Anthropic has made its automated code reviewer generally available, with up to $50,000 in credits for enterprise customers. Organisations that keep their code in Europe cannot claim any of them.
该公司在一篇帖子中表示，Code Review 周四对 Claude Team 和 Enterprise 客户退出研究预览阶段。大多数审查现在在六到十分钟内完成，预览期间约为二十分钟。
Code Review left research preview on Thursday for Claude Team and Enterprise, the company said in a post. Most reviews now finish in six to ten minutes, against about twenty during the preview.
管理员将每个仓库设置为快速、标准或深度模式，安装 GitHub 应用并选择要覆盖的仓库。审查会在新拉取请求上自动运行。
Administrators set each repository to Quick, Standard or Deep, install a GitHub App and pick which repositories to cover. Reviews run automatically on new pull requests.
符合条件的企业可领取 1 万至 5 万美元不等的积分，金额根据其活跃座位数确定。积分自领取之日起有效 30 天，领取窗口将于 1 月 15 日关闭。
Eligible enterprises can claim between $10,000 and $50,000, with the amount fixed by their active seat count. The credits run for 30 days from the claim and the window shuts on 15 January.
排除条款涉及欧洲。连接带有数据驻留的 GitHub Enterprise Cloud 实例（地址为 .ghe.com）的组织无法领取。
The exclusions are where Europe appears. An organisation that connects a GitHub Enterprise Cloud instance with data residency, on a .ghe.com address, cannot claim.
在领取后连接此类实例，整个组织的积分将停止，包括在 github.com 上运行的审查。零数据保留的企业根本无法使用该产品。
Connect one after claiming and the credits stop for the whole organisation, including reviews running on github.com. Firms with zero data retention cannot use the product at all.
该 GitHub 应用拥有其审查仓库的读取权限。六月，一名研究人员展示了 Anthropic 自身的 Claude Code 操作中存在一个提示注入漏洞，该漏洞会泄露具有写入权限的安装令牌。
The GitHub App takes read access to the repositories it reviews. In June a researcher showed that Anthropic’s own Claude Code action carried a prompt injection flaw handing over an installation token with write access.
Anthropic 在四天内修复了核心绕过方法，将问题评为 7.8 分，并支付了 4800 美元赏金。
Anthropic fixed the core bypass within four days, rated the issues 7.8 and paid a $4,800 bounty.
竞争对手在欧洲已更进一步。CodeRabbit 由伦敦的 Atomico 共同牵头筹集了 1.43 亿美元，每周为包括宝马和 Adyen 在内的 1.7 万客户运行两百万次审查，并计划将其欧洲员工人数翻倍。
The competition is further into Europe already. CodeRabbit raised $143M co-led by London’s Atomico, runs two million reviews a week for 17,000 customers including BMW and Adyen, and is doubling its European headcount.
Anthropic 的性能数据来自其自身。该公司称，在 Code Review 之前，16% 的拉取请求收到了实质性评论，之后这一比例上升至 54%，且工程师标记的错误低于 1%。
Anthropic’s performance figures are its own. It says 16% of its pull requests drew substantive comments before Code Review and 54% after, and that engineers marked under 1% incorrect.
该公司还表示，Claude 现在编写了其自身 80% 的代码，这也是审查成为该产品旨在消除的瓶颈的原因。
The company also says Claude now writes 80% of its own code, which is how review became the bottleneck this product was built to clear.
上述内容均未经过独立验证。Anthropic 的文档称，该系统可能会遗漏漏洞或安全缺陷，且审查干净并不代表代码可靠。
None of it has been independently tested. Anthropic’s documentation says the system may miss bugs or security flaws, and that a clean review does not mean the code is sound.
成本是另一个悬而未决的问题。预览期间，每次审查的费用为 15 至 25 美元，而 Anthropic 在合同代币用完后将取消企业折扣。CodeRabbit 首席执行官表示，OpenAI 现在是其主要供应商，而六个月前还是 Anthropic。
Cost is the other open question. Reviews ran $15 to $25 each during the preview, and Anthropic ends enterprise discounts once contracted tokens run out. CodeRabbit’s chief executive says OpenAI is now his main supplier, up from Anthropic six months ago.