Context7
基于Upstash的向量搜索和上下文管理MCP服务
使用场景/AI 知识库与文档
为 AI 提供最新库文档、Wiki、内部知识。减少幻觉,适合写代码、学新框架。
共匹配 1,303 个资源 · 第 3 / 28 页
知识库类 MCP 解决大模型最大的短板——知识过期与幻觉:Context7 按库与版本实时拉取官方文档和代码示例,DeepWiki 把任意 GitHub 仓库编译成可问答的 Wiki,自建方向则常用向量库 / 知识图谱 MCP 接入团队内部资料。
写代码时的典型链路:在 Cursor 里问「Next.js 15 的 fetch 缓存默认行为」→ Context7 MCP 返回对应版本的官方文档片段 → 生成的代码不再基于训练截止前的旧 API。升级大版本、试用新框架时收益最明显。
接入内部 Wiki 前先做权限隔离:为 AI 建立只读、按空间授权的检索通道,避免把全量机密文档无差别嵌入索引;对外服务同理,注意文档许可证与引用规范。
v1.0.4
19 rules over the user documentation that ships with your product: support-period end date, vulnerab
v1.5.1
MCP server for Yandex Wiki with full-text search.
v0.2.5
搜索维基百科、阅读摘要与全文、定位小节、查找相近页面并列出语言版本。
v0.4.1
对接 iNavi Maps API 的 MCP 服务器,为 Claude AI 提供地理编码、POI 检索与路径规划。
v1.0.0
Search and read Financier's developer documentation. No credentials needed.
v0.1.6
CometChat docs search + implementation bundles: add chat, voice, video & moderation to your app.
v2.8.7
Persistent, evidence-backed project knowledge and task context for coding agents.
v1.0.0
Read-only Kyntrax definitions, claims, evidence status and public provenance.
v1.0.0
Travel guides from Wikivoyage. No key required.
v1.0.0
Wikipedia featured content for a day. No key required.
v1.0.0
Species taxonomy from Wikispecies. No key required.
v1.0.0
Use this MCP server to searching and reading public-domain texts on Wikisource.
v1.0.0
Search famous quotes from Wikiquote. No key required.
v1.0.0
Wikipedia page view statistics. No key required.
v1.0.0
Search Wikipedia, fetch article summaries.
v1.0.0
News articles from Wikinews. No key required.
v1.0.0
Wikimedia Commons file search.
v1.0.0
Wikidata entity search, lookup, and bounded SPARQL.
v1.0.0
Open textbooks and manuals from Wikibooks. No key required.
v1.0.0
Use this MCP server to the Tetris wiki, covering every game, piece.
v1.0.0
Keyless MCP server for the Stardew Valley Wiki: search crops, villagers, quests, items.
v1.0.1
Use this MCP server to the Doom Wiki.
v1.0.0
Read-only mdprint product information, documentation and Markdown examples. No document uploads.
v1.0.0
I do everything related to Google Docs
v0.1.19
搜索并获取 Wikidata 实体、执行 SPARQL 查询、解析外部标识符。
v1.0.0
classifier.dev documentation as tools: list, read and search the reference, benchmark and guides.
v1.1.1
Wikimedia Enterprise API docs: search the guides, read every endpoint. Read-only, no account needed.
v26.9.1
GroupDocs.Comparison 的 MCP 服务器——由 AI 智能体比对、分析和检查文档。
v0.22.1
Credit notes against your invoices and purchase orders to suppliers, with VAT and A4 PDFs.
v0.19.13
Family Office Knowledge Graph: read-only MCP door over the public record. Agentic KG Holdings.
v0.1.6
Read-only offline MCP server over a curated Shopware developer knowledge base
v1.2.4
The MCP for the beauty, cosmetics and personal care sector
v0.22.0
Credit notes against your invoices and purchase orders to suppliers, with VAT and A4 PDFs.
v1.0.0
Anonymous read and search over the published nowyourlink developer documentation.
v0.1.1
Multiple Gmail accounts, editable Google Sheets & Docs for AI agents. Deny-by-default access rules.
v1.0.7
面向 AI 智能体的零知识上下文与神经交接能力,AI 记忆的信任层。
v0.1.0
Search the ByJG open source documentation: PHP, DevOps, Helm, JavaScript and AI projects.
v2.0.0
Hybrid search, schema introspection and record read/write over FoxNose collections
v1.0.0
Public PulseWatch offers and monitoring guides. No private customer data or monitor controls.
v1.0.0
Public Upravio contract-management guides and capabilities. No private workspace access.
v1.0.0
Search and retrieve public INMAR services, projects and business information.
v1.0.0
Public Croatian accounting guides and RIFK business information. No private client data.
v1.1.3
Search Wind UI, Magic, Dusk, Telescope and Artisan documentation. Hosted, public, no auth.
v0.1.0
Effect 官方文档的中文译文检索(234 页,v3 + v4)。每条结果带可核验引用:能点回原文小节、能看到该译文对着哪次上游提交译的。零依赖、离线自包含。
v2.41.6
以 Markdown 为主的 Notion API MCP 服务器,含 8 个组合工具与 39 个动作。
v0.2.0
Read-only MCP server for official Tauri 2 documentation snapshots.
v1.0.2
Search and explore source-backed SEO, SaaS, and LLM-wiki knowledge from the XKnow knowledge base.
v1.0.0
Wikidata Q-id shape, value discarded
Context7 以库名+版本精确取文档切片并结构化给模型,省去抓取清洗、上下文更聚焦;直接搜索适合无固定库的泛调研。
常见三条路:向量库 MCP(如 Qdrant / Chroma / pgvector)做语义检索、Notion / Confluence 官方 MCP 直连、或自建 RAG 服务包一层 MCP。按数据敏感级别选。
会增加一次工具调用的延迟(通常几百毫秒),但换来更少返工。控制返回片段大小、只挂载需要的知识库即可。