← Research

DevvDevv

Devv AI

免费+订阅Free + subscription 开发者搜索Developer search 官方文档检索Official documentation search 代码问答Code Q&A 需网络环境Network access required partial lifecycle: active 核验Checked 2026-10-22

Devv 是开发者向的 AI 搜索问答：代码问题→实时抓官方文档→带链接答案；它答代码问题但不改代码库——与 coding 大类的边界就在这条线上。

A developer-facing AI search Q&A: coding questions → live fetch of official docs → answers with links; it answers code questions but does not touch your codebase — that is the boundary with the coding category.

适合与不适合When to use it, when not to

适合：API 用法/报错/库文档速查、带官方链接的答案、中文开发者。不适合：改代码库（归 coding）、粘贴敏感代码不假思索。

Looking up API usage/errors/library docs (answers carry official-doc links), Chinese developers who want reachable search Q&A. Not for: modifying codebases (that is the coding category), pasting sensitive code without thought.

中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-22

出品属地Origin: 国产domestic　Devv AI（团队叙事为中国背景团队出海产品；注册主体未核验到官方声明——诚实标注）。Devv AI (team narrative: a China-background team going overseas; the registered entity was not verified — honest labeling).
国内使用Direct access: 直连不可达direct blocked　devv.ai 大陆直连实测 000（2026-10-22，代理下 200 可达）——直连不可达，需网络环境。devv.ai measured 000 direct from mainland (2026-10-22; 200 via proxy) — unreachable without a network path.
备案状态CAC filing: 备案未核验filing unknown：未核验到对应生成式 AI 服务备案（2026-10-22 检索）；主站面向海外，「未核验到」≠「未备案」——若国内线存在则口径另论。no matching generative-AI service filing found (searched 2026-10-22); the main site targets overseas users — "not found" ≠ "not filed"; a domestic line, if any, is assessed separately.

备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准（按批次滚动更新，本站只抄编号不解读）；「大陆可直连」是本站从大陆网络直连实测的结果，不是厂商承诺。GitHub 类站点当日实测有波动，口径写在依据里。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise. GitHub-hosted sites fluctuated on the measurement day — the note says so.

固定坐标系Fixed coordinate system

8 个维度dimensions, 与同分区其他对象逐项可比each comparable item by item against others in the same section.

模型与开放条件Model & openness: 多模型路由（界面可选多家前沿模型，官方产品面）+ 自建代码文档检索索引（产品差异点：实时抓官方文档）；具体模型清单未逐条清点。
Multi-model routing (several frontier models selectable in the UI, official product surface) + a self-built code-docs retrieval index (the product's differentiator: fetching official docs in real time); the concrete model list not inventoried item by item.
运行位置Where it runs: 网页（devv.ai）+ 浏览器扩展（官网所见）；检索与生成在厂商云端。
Web (devv.ai) + browser extension (seen on the official site); retrieval and generation on the vendor's cloud.
本地文件Local files: 对话历史存账户侧云端；未见官方声明的本机数据面或离线模式（未逐条核验）。
Conversation history stored on the account-side cloud; no officially stated local data surface or offline mode seen (not verified item by item).
关机后的任务Tasks after shutdown: 问答即时返回；未见官方任务队列或「关机续跑」类后台任务面（未逐条核验）。
Q&A returns instantly; no official task queue or "keep-running-after-shutdown" style background task surface seen (not verified item by item).
工具与扩展Tools & extensions: 开发者向 AI 搜索是主轴：代码问题→实时检索官方文档/代码示例→带链接答案；代码库理解问答（官网叙事）；与 coding agent 的区别：它答代码问题，不改你的代码库（边界裁决「搜索向对话归 research」）。
Developer-oriented AI search is the main axis: code questions → real-time retrieval of official docs/code examples → answers with links; codebase-understanding Q&A (official-site narrative); the difference from coding agents: it answers code questions, it does not modify your codebase (boundary ruling: "search-style conversation belongs to research").
上下文与记忆Context & memory: 以对话线程组织检索历史；跨会话记忆未核验到官方声明（未逐条核验）。
Retrieval history organized as conversation threads; cross-session memory not verified to an official statement (not verified item by item).
权限与限制Permissions & limits: 账号+订阅体系（D9：价格不采）；训练用途开关未逐条核验—— 开发者粘贴的代码片段可能含敏感信息，上传前要想清楚。
Account + subscription system (D9: pricing not collected); the training-use switch not verified item by item — code snippets pasted by developers may contain sensitive information; think it through before uploading.
适合什么任务Best for: 适合：查 API 用法/报错/库文档（答案带官方文档链接）、中文开发者要直连可用的搜索问答。不适合：改代码库（那是 coding 大类）、粘贴敏感代码不假思索。
Fits: looking up API usage/errors/library docs (answers carry official-doc links), Chinese developers wanting a directly reachable search Q&A. Does not fit: modifying codebases (that is the coding category), pasting sensitive code without a second thought.

头号误解Top misconception

Devv 是「代码问答搜索」不是 coding agent——它不改你的仓库，别按 Cursor 类预期用它Devv is "code Q&A search", not a coding agent — it does not touch your repository; do not use it with Cursor-class expectations
答案带的文档链接质量取决于实时抓取——付费库/登录墙后的文档抓不到，链接可能到不了正文The quality of doc links in answers depends on live fetching — docs behind paywalls or login walls cannot be fetched and links may not reach the actual content
devv.ai 大陆直连实测 000（2026-10-22，代理 200）——「国产团队但直连不可达」的组合要如实标注devv.ai measured 000 direct from the mainland (2026-10-22, 200 via proxy) — the "domestic team but not directly reachable" combination is noted honestly

未知项清单Known unknowns

注册主体与备案状态（见 cn 块——主体未清）
当前可选模型清单（未逐条清点）
协议专页 URL（站内协议线未能定位直抓——如实标注）
训练用途开关（未逐条核验）

The registered entity and filing status (see the cn block — entity unclear)
The current selectable-model list (not inventoried item by item)
The dedicated agreement-page URL (the in-site agreement line could not be located by direct fetch — honestly noted)
The training-use switch (not verified item by item)

证据来源Evidence sources

判断可回到以下一手源复核。本站核验日Every judgement can be rechecked against the primary sources below. Verified on 2026-10-22，内容更新日content updated on 2026-10-22。

类型Type	名称Name	链接Link
official	Devv · 官网（产品与能力叙事）Devv · official site (product and capability narrative)	https://devv.ai
official	Devv · 官网中文入口（产品面直抓核验 2026-10-22）Devv · official site Chinese entry (product surface direct-fetch verified 2026-10-22)	https://devv.ai/zh
changelog	Devv · 更新日志（官方更新线，changelog）Devv · changelog (official update line, changelog)	https://devv.ai/zh/changelog

实测记录Measured runs

本档未跑检索质量评测；结论只覆盖产品能力与来源披露。

实测建议（三条，都指向官方可读页没给的东西）：

对新版框架文档（发布 <1 月）的索引新鲜度
报错栈粘贴→答案的命中率（对比直接搜 Stack Overflow）
免费档与订阅档的模型切换差异

This site has not completed measured runs. The coding-oriented test protocol is at protocols/coding/_protocol.md (the coding task suite does not apply to creation products).

Suggested measurements (three, all pointing at things the official readable pages do not give):

Index freshness for brand-new framework docs (released <1 month ago)
Hit rate of pasted error-stack → answer (vs searching Stack Overflow directly)
Model-switching differences between the free tier and the paid tier