1Password 官方 Environments MCP server（本地 stdio，随桌面应用）：凭据类 MCP 的1Password official Environments MCP server (local stdio, rides
部分核验Partly verified
详情 →Details →
AI 工具AI Tools · MCP
MCP 工具MCP tools
MCP server 的权限范围、传输方式与输出可用性。收录标准：已发布为可安装的 MCP server、有官方仓库或文档可回溯、近 30 天有实质更新。这里是能力缺口查表，不是选型对比。The permission scope, transport and output usability of MCP servers. Admission criteria: released as an installable MCP server, has an official repository or documentation that can be traced back, and has substantive updates within the last 30 days. This is a capability-gap reference table, not a selection comparison.
- 本站按 10 大类收录 AI 产品。编程、对话、图像、视频、音频、研究、自动化、框架、MCP 工具、本地运行器——每类都有明确的收录标准，写在各大类的目录页里。This site files AI products into 10 categories. Coding, chat, image, video, audio, research, automation, frameworks, MCP tools and local runners — each with explicit admission criteria, written on the category page itself.
- 大类看目录，形态看字段。一个产品属于哪一类由收录边界决定；它怎么装进你的工作流（IDE / CLI / 厂商云 / 框架 / MCP…）是另一根独立的轴。Category comes from the shelf, form from a field. Which category a product belongs to is decided by admission boundaries; how it installs into your workflow (IDE / CLI / vendor cloud / framework / MCP…) is a separate axis.
- 每条档案都是核验快照。字段逐条挂官方出处与核验日期，本站不实测、不排名，未知就写未知。Every entry is a verification snapshot. Each field carries an official source and a checked date; we run no benchmarks, publish no rankings, and write "unknown" when it is unknown.
深水指南Deep guides
产品档案回答「这个工具是什么」；深水指南回答「这类工程问题怎么办」。六篇各管一件事：编排、记忆、规划、评测、成本、护栏——讲机制与取舍，不做框架排名。每篇的关键事实都挂官方文档或论文出处。Product entries answer "what is this tool"; the deep guides answer "how do you handle this class of engineering problem". Six guides, one topic each — orchestration, memory, planning, evaluation, cost, guardrails — covering mechanisms and trade-offs, with no framework rankings. Every key fact carries an official-doc or paper source.
规划Planning
学术研究工具地图A map of academic research tools
按「找论文 / 读论文 / 查引用态度 / 做综述」四件事分组，给出 Semantic Scholar、Scite、SciSpace 等工具的边界与官方入口。Grouped by four jobs — find papers, read papers, check citation stance, build a review — with the boundaries and official links for Semantic Scholar, Scite, SciSpace and others.
记忆系统Agent memoryagent 记忆系统Agent memory systems
会话内的状态、跨会话的持久事实、外部记忆库——三层记忆各自解决什么问题，以及共同的坑。In-session state, cross-session facts, and external stores — what each memory layer solves, and the pitfalls they share.
规划PlanningAPI 还是本地跑API or local inference
四条轴决定：成本结构、数据与合规、延迟与吞吐、模型能力与更新；用一张决策表缩小到一条路线。Four axes decide it — cost structure, data and compliance, latency and throughput, model capability and updates. A decision table narrows it to one route.
规划Planning2026 Q3–Q4 变更速记：10 条带日期的官方变更Q3–Q4 2026 change digest: 10 dated official changes
把「刚上手就过时」做成可回看的季度页：10 条官方发布、停运与迁移，每条带日期、影响与来源；已宣布未生效的单独标注。A quarterly page you can revisit instead of guessing what changed: 10 official launches, retirements and migrations, each with a date, impact and source, with announced-but-not-yet-effective items marked separately.
规划Planning聊天助手怎么选How to choose a chat assistant
不排名、不报价格：先定中文、数据、价格、生态四个约束，再按场景缩小到 2–3 个候选。No rankings and no prices: fix four constraints — Chinese, data, price, ecosystem — then narrow to 2–3 candidates by scenario.
护栏与安全Guardrails & safety采购与权限审计清单：问厂商的 18 个问题Procurement and access audit checklist
一份可直接发给厂商的采购问卷，覆盖身份、数据、日志、模型、合同与退出六组；附权限矩阵模板与 6 项季度审计，问题都对应官方文档可查项。A vendor questionnaire you can send as-is, covering identity, data, logs, model, contract and exit in six groups, with a permission-matrix template and six quarterly audit items, each answerable from official docs.
护栏与安全Guardrails & safety国内可用性怎么读：直连、主体与备案的口径How to read mainland availability
用 tools 档案里的 cn 三个字段回答「国内能不能连、谁在运营、有没有备案」；给出 225 份档案的统计口径、限制与自查方法，不做实时可用性承诺。Three cn fields in the tool dossiers answer whether a product connects from the mainland, who operates it and whether it is filed; this page gives the counting rules, limits and a self-check method for 225 dossiers, with no real-time guarantee.
规划PlanningAgent 框架：LangGraph / OpenAI Agents SDK / Google ADK / Claude Agent SDK 怎么选Agent frameworks: how to choose
按状态与持久化、人在回路、模型绑定与部署形态四面选；低层编排与开箱 harness 是两类东西。Choose by state and persistence, human-in-the-loop, model binding and deployment; low-level orchestration and batteries-included harnesses are different classes.
规划Planning浏览器 Agent 三选一：Playwright MCP / Browserbase MCP / Chrome DevTools MCPBrowser agents: Playwright, Browserbase, DevTools
三个浏览器 MCP 的适用条件、能力边界与价格档位；按「本地还是云端、要不要模型、要不要性能分析」给出一句选择，并写明什么情况下换另一个。Applicability, capability limits and plan tiers for three browser MCP servers; a one-line choice based on local vs cloud, whether a model is needed and whether performance analysis is required, plus when to switch.
规划Planning聊天助手：ChatGPT / Gemini / Claude / Copilot 怎么选Chat assistants: ChatGPT vs Gemini vs Claude vs Copilot
按数据条款、研究线、生态绑定与大陆可达四项硬约束筛选，再按场景落到一个主助手加一个备选。Filter by data terms, research line, ecosystem lock-in and mainland reachability, then pick one primary assistant and one backup by scenario.
规划Planning写代码：终端 / IDE / 云端 Agent 怎么选Coding agents: terminal vs IDE vs cloud
先定执行面（终端 / IDE / 云端），再比权限模型、后台续跑与仓库上下文；给每个候选写明适用条件与一句「什么情况下选另一个」。Fix the execution surface first (terminal, IDE, cloud), then compare permission models, background continuation and repo context; each candidate gets its conditions and one line on when to pick another.
规划Planning本地跑模型：Ollama / LM Studio / llama.cpp / vLLM 怎么选Local runners: Ollama vs LM Studio vs llama.cpp vs vLLM
按硬件、量化格式、并发与服务化四面选 runtime；桌面图形、单机脚本、生产服务是三条不同路线。Choose by hardware, quantization format, concurrency and serving needs; desktop GUI, single-machine scripting and production serving are three different routes.
规划Planning图像与视频：一条内容该用哪条工具链Image and video pipelines
先分图像与视频两条链，再按输入模态、权利条款与工作流位置选；每个候选写明适用条件与替代路线。Split image and video into two chains, then choose by input modality, rights terms and workflow position; each candidate gets its conditions and an alternative.
规划Planning研究工具：Perplexity / Kagi / NotebookLM / Elicit 怎么选Research tools compared
按检索深度、来源透明度、资料锚定与引用管理四面选；搜索引擎、资料集问答与文献抽取是三件事。Choose by search depth, source transparency, source grounding and citation management; search engines, source-set Q&A and literature extraction are three different jobs.
成本控制Cost control成本算式：云端、Agent、本地与视频四条线Cost formulas: cloud, agents, local and video
不给价格数字，只给四条可复算的算式与口径：token、席位、显存电费、视频秒数；每条附官方定价页指针。No price figures, only four reproducible formulas and conventions — tokens, seats, local power, video seconds — each with an official pricing pointer.
成本控制Cost controlagent 的成本控制Cost control for agents
钱花在哪、怎么省：token 记账、前缀缓存、上下文膨胀与步数上限——只讲机制，不报价格。Where the money goes and how to curb it: token accounting, prefix caching, context bloat, and step caps — mechanisms only, no prices.
评测Evaluationagent 评测Evaluating agents
先定义任务和成功条件，再用固定样本检查改动、记录实际失败；评测器的判断也需要核对。No evaluation, no iteration — offline regression testing, live monitoring, and the pitfalls of evaluators themselves.
规划Planning练习包 · Builder：用 10 篇站内文档搭一个带引用的问答Exercise: cited answers from 10 source pages
一个可独立完成的 RAG 练习：选 10 篇站内页面做语料，产出 10 题答案、引用表、评测表与失败记录；附评分标准与参考答案结构。A self-contained RAG exercise: use 10 site pages as a corpus and produce 10 answers, a citation table, an evaluation sheet and a failure log, with a rubric and a model-answer outline.
规划Planning练习包 · Creator：把一份官方资料做成三件套并留下事实表Exercise: three assets from one official source
选一份官方产品页，产出短文、图像提示词与 30 秒口播稿，并附 5 条事实核对表与 2 条被拒的说法；评分标准看引用、边界与一致性。Take one official product page and produce a short post, an image prompt and a 30-second script, plus a five-claim fact table and two rejected claims; the rubric checks citations, boundaries and consistency.
规划Planning练习包 · Manager：写一页供应商决策备忘录Exercise: a one-page vendor decision memo
用三个成本维度对比「调 API」与「自己部署」，交一页决策备忘录、评分表、5 个尽职调查问题与退出计划；评分标准看口径、可复核与退出路径。Compare API usage with self-hosting on three cost dimensions and deliver a one-page decision memo, a score sheet, five due-diligence questions and an exit plan; the rubric checks conventions, reproducibility and exit paths.
规划Planning失败库：十个真实误判与证据Failure library: ten real misreads with evidence
从 225 份档案的 pitfalls 汇总十个高频误判：现象、为什么会踩、怎么避免，每条附档案链接与核验日。Ten recurring misreads compiled from the pitfalls across 225 profiles — symptom, cause, avoidance, with a profile link and verification date for each.
护栏与安全Guardrails & safetyagent 的护栏Guardrails for agents
输入输出与工具三层防线、触发即止的 tripwire 机制，以及并行与阻塞的成本取舍。Input, output, and tool-level defenses, tripwires that halt on trigger, and the cost trade-off between parallel and blocking checks.
规划Planning图像 / 视频 / 音频工具怎么选How to choose image, video and audio tools
商用许可先过筛，再看输入输出与编辑链、本地还是云端、中文与区域可用；最后按产量估算成本。Screen for commercial licensing first, then input/output and edit chains, local vs cloud, Chinese and regional access; finally estimate cost by volume.
多 agent 编排Multi-agent orchestration多 agent 编排Multi-agent orchestration
多个 agent 怎么分工、怎么交接、怎么收拢结果，以及什么时候根本不该上多 agent。How multiple agents divide work, hand off, and merge results — and when not to go multi-agent at all.
规划Planningagent 的规划Planning for agents
先想清楚步骤再动手，还是边做边定——规划模式怎么选，以及规划为什么会失效、失效了怎么救。Plan first or decide as you go — how to choose a planning pattern, and why plans fail.
规划Planning任务：做一条 60 秒产品视频Playbook: producing a 60-second product video
脚本、分镜、生成、剪辑、权利登记五段；含单段时长限制、成本算式、失败模式与交付清单。Five stages — script, storyboard, generation, editing and rights logging — with per-clip limits, a cost formula, failure modes and a delivery checklist.
规划Planning任务：用编码 Agent 修一个真实 bugPlaybook: fixing a real bug with a coding agent
从仓库上下文、复现、计划、改动、测试到审查的六步流程；含权限设置、成本算式、失败模式与验收单。A six-step flow from repo context and reproduction through planning, editing and testing to review, with permission settings, a cost formula, failure modes and an acceptance list.
规划Planning任务：在自己的机器上跑一个模型Playbook: running a model on your own machine
显存、量化档、runtime、评估、监控五步；含体积算式、量化索引、失败模式与验收单。Five steps — VRAM, quantization tier, runtime, evaluation, monitoring — with a size formula, the quantization index, failure modes and an acceptance list.
规划Planning任务：做一份可核查的研究简报Playbook: producing a checkable research brief
问题、检索、来源、综合、复核五步；每个结论挂来源与日期，含工具分工、失败模式与验收单。Five steps — question, search, sources, synthesis, review — with a source and date on every claim, plus tool roles, failure modes and an acceptance list.
规划Planning任务：搭一个客服 Agent 并上线Playbook: building and shipping a customer-service agent
意图、知识库、工具、护栏、评估、上线六段；含生态锁定、许可、人工兜底、验收单与失败模式。Six stages — intents, knowledge base, tools, guardrails, evaluation, launch — with ecosystem lock-in, licensing, human fallback, an acceptance list and failure modes.
规划Planning团队 90 天落地 playbook：从试点到权限收口A 90-day team rollout playbook
把团队引入 AI 拆成 0–30、31–60、61–90 三段，每段给目标、动作、验收点与负责人；权限矩阵与审计项按厂商官方 admin 文档落地，不写空泛口号。A 0–30, 31–60, 61–90 rollout with goals, actions, acceptance points and owners per phase; the permission matrix and audit items follow vendor admin documentation rather than slogans.
规划Planning20 题自测：AI 基础、评测与选型20-question self-test: fundamentals and evaluation
20 道只考站内已核验内容的选择/简答题，附参考答案与评分标准；先做题再看答案，36 分以上算优秀。Twenty questions drawn only from verified on-site content, with an answer key and scoring rules; do the questions first, then check the answers, 36+ is excellent.
规划Planning工具箱：Agent 评估、护栏、客服验收、故障清单Toolkit: agent ops checklists
四个可直接复制的文本块：评估 rubric、四层护栏清单、客服验收单、故障复盘清单。Four copyable text blocks: an evaluation rubric, a four-layer guardrail checklist, a service acceptance sheet and an incident review checklist.
规划Planning工具箱：提示词、上下文预算、RAG 清单、选型表Toolkit: prompt and context
四个可直接复制的文本块：提示词结构、上下文预算表、RAG 分块与重排清单、模型选型工作表。Four copyable text blocks: a prompt structure, a context budget table, a RAG chunking/reranking checklist and a model selection worksheet.
规划Planning工具箱：权限矩阵、许可清单、视频权利表、成本算式Toolkit: rights and cost
四个可直接复制的文本块：Agent 权限矩阵、模型许可检查表、视频版权登记表、成本算式模板。Four copyable text blocks: an agent permission matrix, a model licence checklist, a video rights log and a cost formula template.
规划Planning工具箱：四张工作表（立项、发布、迁移、泄露）Toolkit: four worksheets
四个可直接复制的检查表：项目立项、内容发布前、模型迁移评估、数据泄露路径排查。Four copyable checklists: project kickoff, pre-publish, model migration assessment and data-leak path review.
先回答一件事：你到底缺什么能力Answer one question first: what capability are you actually missing
MCP 不是「让 AI 更聪明」，是给 AI 接上它本来够不到的东西。如果你要的只是聊天和写代码，你不需要它。你装，它调——这一大类 32 份档案，按出身分：7 份是 MCP 官方 reference server，25 份由各自厂商或官方团队维护（Playwright、Context7、GitHub 以及 Neon / Notion / Stripe 等）。每份档案的出身与权限范围写在 scope 维里。MCP is not "making AI smarter" — it is giving AI access to what it could not otherwise reach. If all you want is chat and writing code, you do not need it. You install it, the AI calls it — 32 entries by origin: 7 are the MCP org's reference servers, and 25 are maintained by their own vendor or official teams (Playwright, Context7, GitHub, and Neon / Notion / Stripe and others). Each entry states its origin and permission scope in the scope axis.
-
1我要读写本地文件 / 操作 GitI need to read/write local files / work with Gitfilesystem、gitfilesystem, git两个都是读写权限，git 那一类值得先看 scope 里能限制到哪些目录。Both have read and write permissions; for the git category it is worth checking what directories the scope can be limited to.
-
2我要它记住东西、跨会话召回I need it to remember things across sessionsmemorymemory这是最容易被误解的一类——它不是「模型记忆」，是外部知识图谱存储。This is the most misunderstood category — it is not "model memory" but an external knowledge-graph store.
-
3我要让 AI 上网拿资料I need AI to reach the web for materialcontext7、fetchcontext7, fetch风险最高的一类是 fetch：档案里引了官方警告——可访问本地与内网 IP，无地址白名单、无审批机制。fetch carries the highest risk: the entry quotes an official warning — it can reach local and internal IPs, with no address allowlist and no approval mechanism.
-
4我要控制真实浏览器I need to drive a real browserplaywrightplaywright注意官方明确声明来源限制不是安全边界，需要真隔离得用
--isolated或容器化网络策略。Note the official explicit statement that source restrictions are not a security boundary; real isolation requires `--isolated` or a containerised network policy. -
5我要点时间 / 拆解推理 / 跨平台搜文件I want time / decomposition / cross-platform file searchtime、sequential-thinking、everythingtime, sequential-thinking, everything三个都是只读，风险最低。All three are read-only, so risk is lowest.
-
6我要操作 GitHub（仓库 / Issue / PR / Actions）I need to work with GitHub (repos / issues / PRs / Actions)看 github-mcpSee github-mcpGitHub 官方自营的 MCP server，与归档的旧参考 server 不是同一条线；权限由 GitHub token 决定，可收紧到只读。GitHub's own MCP server — not the same line as the archived reference server; permissions follow the GitHub token and can be tightened to read-only.
⚠ 一个容易忽略的维度：数据出境。Context7 默认以远程服务交付，查询内容会发到 Upstash 服务器；GitHub MCP Server 提供远程托管与本地 Docker 两种形态；其余八个都是本地 stdio 进程。另外——装 MCP 需要你手动做，调用不需要：server 启动后自动把工具清单报给你的客户端，之后调不调、什么时候调，由 agent 自己判断，你只在审批环节介入。⚠ An easily overlooked dimension: data leaving the country. Context7 is delivered as a remote service by default — queries go to Upstash's servers; GitHub MCP Server offers both a remote hosted and a local Docker form; the other eight are local stdio processes. And — installing an MCP server is manual; calling it is not: once running, the server reports its tool list to your client automatically, and after that the agent decides whether and when to call it. You only get involved at the approval step.
全部对象All objects
按统一坐标系排列。点开任一对象可看到 模型与开放条件 / 运行位置 / 本地文件 / 关机后的任务 / 工具与扩展 / 上下文与记忆 / 权限与限制 / 适合什么任务 / 传输方式 / 认证机制 / 权限范围 的完整记录、证据链接与未知项清单。Everything on one coordinate system. Open any object to see its full record across Model & openness / Where it runs / Local files / Tasks after shutdown / Tools & extensions / Context & memory / Permissions & limits / Best for / Transport / Authentication / Permission scope, with source links and a list of unknowns.
Atlassian 官方 Remote MCP server：一个端点接 Jira+Confluence 两面——agentAtlassian's official Remote MCP server: one endpoint covers Ji
部分核验Partly verified
详情 →Details →
Brave 官方搜索 MCP server（MIT）：给 agent 补一路独立检索源（web/新闻/图像/地点）。 一条归Brave's official search MCP server (MIT): an independent retri
部分核验Partly verified
详情 →Details →
Browserbase MCP ServerBrowserbase MCP Server
Browserbase
Browserbase 官方 MCP server：云端浏览器 + Stagehand 自然语言网页驱动，6 工具 （staBrowserbase official MCP server: cloud browser + Stagehand nat
部分核验Partly verified
详情 →Details →
Chrome DevTools MCPChrome DevTools MCP
Google（Chrome DevTools 团队）Google (Chrome DevTools team)
Chrome 官方团队做的「让 agent 看见浏览器」：CDP 直连，导航、交互、截图、performance traceChrome's own team built 'let the agent see the browser': direc
部分核验Partly verified
详情 →Details →
Cloudflare MCP ServerCloudflare MCP Server
Cloudflare
Cloudflare 官方 MCP server（Apache-2.0）：agent 写 Worker、部署、管 D1/R2Cloudflare's official MCP server (Apache-2.0): agents write Wo
部分核验Partly verified
详情 →Details →
Context7Context7
Upstash
远程托管的文档服务 —— 为 Agent 提供版本感知的库文档，MIT 开源但核心服务在 Upstash 侧。A remotely hosted documentation service — version-aware librar
部分核验Partly verified
详情 →Details →
Datadog 自营的官方 MCP server（mcp.datadoghq.com/v1/mcp）：agent 排障时直查Datadog's self-run official MCP server (mcp.datadoghq.com/v1/m
部分核验Partly verified
详情 →Details →
Everything MCP ServerEverything MCP Server
Model Context Protocol
不是可用工具——官方明说这是给 MCP 客户端开发者做功能测试的 server，唯一用途是验证客户端协议实现是否正确。Not a usable tool — the official page states plainly that this
已核验Verified
详情 →Details →
Exa 官方 MCP server（MIT）：默认就两工具——web_search_exa 搜索、web_fetch_exaExa's official MCP server (MIT): two default tools — web_searc
部分核验Partly verified
详情 →Details →
Fetch MCP ServerFetch MCP Server
Model Context Protocol
把网页内容转成 Markdown 的 reference server，官方在其 README 里明确标注了安全警告：它能访A reference server that converts web pages to Markdown, with a
已核验Verified
详情 →Details →
Figma 官方 MCP server：把设计系统上下文（组件/变量/Code Connect）喂给编码 agent， 还能Figma's official MCP server: design-system context (components
部分核验Partly verified
详情 →Details →
Filesystem MCP ServerFilesystem MCP Server
Model Context Protocol
MCP 官方 reference server 里的文件操作实现。它是「白名单目录」式访问：必须显式声明允许目录，不允许静默The file-operation implementation among MCP's official referen
已核验Verified
详情 →Details →
Firecrawl MCP ServerFirecrawl MCP Server
Firecrawl
Firecrawl 官方 MCP server（MIT）：web 上下文全家桶——搜索、抓正文、解析 PDF、 站点级 crFirecrawl's official MCP server (MIT): the full web-context ki
部分核验Partly verified
详情 →Details →
Git MCP ServerGit MCP Server
Model Context Protocol
MCP 官方 reference server 里的版本控制实现。它作用域限定在指定 git 仓库，但具备完整的本地写操作能The version-control implementation among MCP's official refere
已核验Verified
详情 →Details →
GitHub 官方自己的 MCP server（仓库自述 "GitHub's official MCP Server"）， GitHub's own official MCP server — not the archived MCP-org re
部分核验Partly verified
详情 →Details →
Grafana MCP ServerGrafana MCP Server
Grafana Labs
Grafana 官方 MCP server（Apache-2.0，Go 实现）：把 agent 接进你的 Grafana 实Grafana's official MCP server (Apache-2.0, Go): plugs agents i
部分核验Partly verified
详情 →Details →
Intercom 官方托管 MCP server（闭源 remote，streamable-http）：客服数据接进 AI Intercom official hosted MCP server (closed-source remote, str
部分核验Partly verified
详情 →Details →
Linear MCPLinear MCP
Linear
Linear 官方 MCP server（mcp.linear.app/mcp）：agent 搜 issue、建 issueLinear's official MCP server (mcp.linear.app/mcp): agents sear
部分核验Partly verified
详情 →Details →
Knowledge Graph Memory ServerKnowledge Graph Memory Server
Model Context Protocol
用本地知识图谱做跨会话持久记忆的 reference server。它记忆落盘、跨会话共享，但没有访问控制。A reference server implementing cross-session persistent memor
已核验Verified
详情 →Details →
MongoDB 官方 MCP server（mongodb-js 仓，Apache-2.0）：agent 连库查集合、跑聚合MongoDB's official MCP server (mongodb-js repo, Apache-2.0): a
部分核验Partly verified
详情 →Details →
Neon 官方 MCP server（MIT）：agent 管 Neon 项目、建分支、跑 SQL。 特色是分支模型——agNeon's official MCP server (MIT): agents manage Neon projects,
部分核验Partly verified
详情 →Details →
Notion MCPNotion MCP
Notion
Notion 官方自己的 MCP server：纯远程托管（mcp.notion.com/mcp），agent 能搜工作区、Notion's own MCP server: fully hosted (mcp.notion.com/mcp) — a
部分核验Partly verified
详情 →Details →
微软维护的浏览器自动化 server。它用无障碍树而非截图驱动，不需要视觉模型。输出是结构化快照。A browser automation server maintained by Microsoft, driven by
已核验Verified
详情 →Details →
Sentry 官方的 agent 工具线：把「查崩溃、看事件、对 release」包成 MCP 工具， 编码 agent 能Sentry's official agent tooling line: crashes, events and rele
部分核验Partly verified
详情 →Details →
Sequential Thinking MCP ServerSequential Thinking MCP Server
Model Context Protocol
提供单个 sequential_thinking 工具，让 Agent 以可修订、可分支的方式逐步推进复杂问题——它不产生智Provides a single sequential_thinking tool that lets an agent
已核验Verified
详情 →Details →
Shopify 官方 Dev MCP server（npm ISC，本地 stdio，无鉴权）：给编码 agent 补 ShShopify official Dev MCP server (npm ISC, local stdio, no auth
部分核验Partly verified
详情 →Details →
Slack MCP ServerSlack MCP Server
Slack（Salesforce）
Slack 自营的官方 MCP server（mcp.slack.com/mcp）：agent 能搜工作区、读线程、代拟消息Slack's self-hosted official MCP server (mcp.slack.com/mcp): a
部分核验Partly verified
详情 →Details →
Stripe 官方给 agent 的开发工具线（stripe/ai，MIT）：查 API 用法、查账户对象、 测试模式试操作Stripe's official agent tooling line (stripe/ai, MIT): look up
部分核验Partly verified
详情 →Details →
Supabase MCPSupabase MCP
Supabase
Supabase 官方 MCP server（Apache-2.0，仓现名 supabase/mcp）：编码 agent 直Supabase's official MCP server (Apache-2.0, repo now supabase/
部分核验Partly verified
详情 →Details →
Time MCP ServerTime MCP Server
Model Context Protocol
只做两件事：查当前时间、换时区。不读写项目文件；仍应核对所用 server 的实现与权限。A reference server providing date and time information to agen
已核验Verified
详情 →Details →
Zapier MCPZapier MCP
Zapier
Zapier MCP 是「千余应用的入口」：你在 Zapier 侧勾选哪些 action，agent 就有哪些工具—— 工具Zapier MCP is an entry point to a thousand apps: whatever acti
部分核验Partly verified
详情 →Details →
本站目前是「核验快照」，还没有实测数据This site is currently a verification snapshot — no measured runs yet
下面每个 server 的权限范围、传输方式与工具清单都能回到官方 README 核对，但我们还没有在真实客户端里跑过它们。Every server's permission scope, transport and tool list below can be traced back to its official README, but we have not yet run them inside a real client. 实测协议已经写好（统一任务、统一验收、记录人工介入与返工次数），任务清单也已就绪，The measurement protocol is written (shared tasks, shared acceptance criteria, recording human interventions and rework counts) and the task list is ready, but it is 尚未执行。not yet run.
MCP server 的风险不在「能不能干活」，而在权限边界是否清楚、越界是否被拒。The risk in an MCP server is not whether it works, but whether the permission boundary is clear and overreach is refused. 真正的选型请用你自己的输入、预算与验收标准跑一遍。For an actual choice, run them yourself against your own inputs, budget and acceptance criteria.
11 个维度，每格都能回溯11 dimensions, every cell traceable
所有对象在同一坐标系下被描述（MCP 服务器另有 3 个特有维度）。All objects are described on the same coordinate system (MCP servers have 3 additional dimensions). 点开任一对象，每一格下面都有官方源链接与核验日期——不认同可以自己回去查。Open any object and every cell carries an official source link and a checked date — disagree and verify it yourself.
01
模型与开放条件Model & openness
用谁的模型？能不能换第三方？多模型路由怎么配？Whose model does it use? Can you switch to a third party? How is multi-model routing configured?
02
运行位置Where it runs
本地 / 云端 / 混合？索引在本地还是云端？远程控制不等于云端执行。Local, cloud or hybrid? Is the index local or remote? Remote control is not the same as cloud execution.
03
本地文件Local files
能访问多大范围？未打开的大文件怎么读？符号级能力如何？How wide an area can it reach? How does it read large files you have not opened? How capable is it at the symbol level?
04
关机后的任务Tasks after shutdown
电脑关了还能继续跑吗？这是本地形态最常见的误解来源。Does it keep running once the computer is off? This is the most common misconception about local forms.
05
工具与扩展Tools & extensions
工具链有多宽：MCP · LSP · 插件 · Hooks · 自定义指令。How wide is the toolchain: MCP · LSP · plugins · hooks · custom instructions?
06
上下文与记忆Context & memory
索引算法、上下文窗口、会话恢复与分叉、跨会话记忆边界。Indexing algorithm, context window, session resume and fork, and the boundary of cross-session memory.
07
权限与限制Permissions & limits
沙箱与审批边界、凭据管理、数据是否出境、额度如何计算。Sandbox and approval boundaries, credential handling, whether data leaves your machine, and how quota is charged.
08
适合什么任务Best for
最擅长什么任务？什么情况下别用它？What tasks is it best at? When should you not use it?
09
传输方式Transport
客户端怎么连上它：stdio / SSE / Streamable HTTP / 远程。How the client connects to it: stdio / SSE / Streamable HTTP / remote.
10
认证机制Authentication
怎么证明身份。注意认证 ≠ 授权，前者是「你是谁」，后者是「你能做什么」。How identity is proved. Note that authentication is not authorisation: the former is who you are, the latter is what you may do.
11
权限范围Permission scope
能碰到多少东西？能改吗？能删吗？能收紧吗？这是 MCP 分区第一优先级。How much can it reach? Can it modify? Delete? Be tightened? This is the first priority in the MCP track.
全部数据一览All data at a glance
同一张表横向对比。点对象名进详情页看逐格证据。The same table side by side. Click an object name to see per-cell evidence.
| 对象Object | 形态Form | 厂商Vendor | 适合Best for | 核验日Checked | 状态Status |
|---|---|---|---|---|---|
| 1P1P1Password Environments MCP Server1Password Environments MCP Server | MCPMCP | 1Password | 适合：让编码 agent（Claude Code/Codex/Cursor/Kiro）管理开发密钥的「环境侧」——建环境、 加变量、挂 .eFits: coding agents managing the environment side of dev secrets (crea | 2026-10-31 | partial |
| AtATAtlassian Remote MCP Server（Jira/Confluence）Atlassian Remote MCP Server（Jira/Confluence） | MCPMCP | Atlassian | 适合：企业研发流里 agent 取 Jira 需求上下文、写 Confluence 文档。 不适合：要写审批流；Data Center 自建Fits: the enterprise 'requirements → code → docs' agent loop. Not a fi | 2026-10-30 | partial |
| BrBRBrave Search MCP ServerBrave Search MCP Server | MCPMCP | Brave | 适合：给 agent 补「现役搜索」能力（reference 的 Brave 版归档后，官方 @brave 包接替）、 要独立于大模型内置搜Fits: an independent retrieval lane besides built-in model search; low | 2026-10-30 | partial |
| BrBRBrowserbase MCP ServerBrowserbase MCP Server | MCPMCP | Browserbase | 适合：要「云端浏览器 + 自然语言驱动网页」的 agent（无需本地起浏览器）；要托管起步、 LLM 成本由厂商包的省事路线。 不适合：要开Fits: cloud browser without local setup, natural-language act/observe/ | 2026-10-31 | partial |
| ChCHChrome DevTools MCPChrome DevTools MCP | MCPMCP | Google（Chrome DevTools 团队）Google (Chrome DevTools team) | 适合：编码 agent 调试真实页面、性能分析、E2E 验证闭环（改代码→看浏览器→再改）。 不适合：要无头批量抓取（那是 fetch/fiFits: the coding agent's 'edit code → verify in browser' loop, perform | 2026-10-30 | partial |
| ClCLCloudflare MCP ServerCloudflare MCP Server | MCPMCP | Cloudflare | 适合：Cloudflare 栈的 agent 部署流（写 Worker→部署→查日志）、docs 检索。 不适合：要「只配置不部署」的硬边界Fits: deploy-debug loops on the Cloudflare stack plugged into agents. | 2026-10-30 | partial |
| C7COContext7Context7 | MCPMCP | Upstash | 需要查当前版本库文档的场景：库升级后的 API 变更、 新版本引入的配置项、模型不确定的第三方库用法。 不适合查内部代码或私有库。Scenarios needing current-version library documentation: API changes a | 2026-10-23 | partial |
| DaDADatadog MCP ServerDatadog MCP Server | MCPMCP | Datadog | 适合：编码/运维 agent 排障时取生产实时上下文（日志/指标/trace/incident）、 在 IDE 里对着真实数据调试。 不适合Fits: coding/ops agents debugging against live production data in the | 2026-10-23 | partial |
| EEVEverything MCP ServerEverything MCP Server | MCPMCP | Model Context Protocol | 仅适合一件事：测试你的 MCP 客户端是否正确实现了协议。 不适合任何实际任务场景。Scenarios where you are testing an MCP client implementation rather th | 2026-10-23 | verified |
| ExEXExa MCP ServerExa MCP Server | MCPMCP | Exa | 适合：给 agent 一路「搜索+读正文」二合一的检索面（默认两工具就覆盖）、 零门槛接入（匿名可用）、要多步研究 Agent 的进阶面。 Fits: a combined search+read retrieval lane, zero-barrier onboarding ( | 2026-10-23 | partial |
| FFEFetch MCP ServerFetch MCP Server | MCPMCP | Model Context Protocol | 把公开网页内容转成 Markdown 供 Agent 读取。 不适合在能访问内网的环境中使用， 也不适合需要登录态的页面。Scenarios needing to pull a web page's content as Markdown for an agen | 2026-10-23 | verified |
| FiFIFigma MCP ServerFigma MCP Server | MCPMCP | Figma | 适合：设计系统驱动的编码流（agent 按真实组件/变量生成代码）、代码↔画布双向流。 不适合：不在 MCP catalog 的客户端（要开Fits: design-system-driven frontend generation, design↔code sync teams | 2026-10-30 | partial |
| FSFIFilesystem MCP ServerFilesystem MCP Server | MCPMCP | Model Context Protocol | 让 Agent 安全读写指定目录内的文本与媒体文件。 不适合需要删除操作或需要跨目录聚合的场景。Scenarios where an agent must read or write within explicitly declared | 2026-10-23 | verified |
| FiFIFirecrawl MCP ServerFirecrawl MCP Server | MCPMCP | Firecrawl | 适合：要「搜索→抓正文→结构化 JSON→浏览器交互→站点级 crawl/map→页面监控」全链一面包齐的 agent； 要 keylessFits: agents wanting the whole chain — search → scrape → structured JS | 2026-10-23 | partial |
| GGIGit MCP ServerGit MCP Server | MCPMCP | Model Context Protocol | 让 Agent 读取仓库状态、历史与 diff，并在受控前提下执行提交与分支操作。 不适合需要 push / PR / 远程仓库管理的场景。Scenarios where an agent should make version-control operations inside | 2026-10-23 | verified |
| GHGIGitHub MCP ServerGitHub MCP Server | MCPMCP | GitHub | 一句话：让任意 MCP 客户端用自然语言操作 GitHub。 客户端包括 VS Code、Claude Desktop、Cursor、WinAnyone who wants their MCP client (VS Code, Claude Desktop, Cursor, Wi | 2026-10-23 | partial |
| GrGRGrafana MCP ServerGrafana MCP Server | MCPMCP | Grafana Labs | 适合：agent 在自托管或 Grafana Cloud 上查仪表盘/指标/告警、改仪表盘、 跨数据源统一 SQL 查询。 不适合：零配置无Fits: agents querying dashboards/metrics/alerts on self-hosted or Graf | 2026-10-23 | partial |
| InINIntercom MCP ServerIntercom MCP Server | MCPMCP | Intercom | 适合：客服/支持团队把 Intercom 数据接进 AI 工具（Claude/Cursor/VS Code 等官方在册客户端）—— 查会话、Fits: support teams plugging Intercom into Claude/Cursor/VS Code and o | 2026-10-31 | partial |
| LiLILinear MCPLinear MCP | MCPMCP | Linear | 适合：编码 agent 干完活顺手建/更新 issue、从 Linear 拉「这个需求当时怎么说的」。 不适合：要审批流的严格写管控；非 LFits: coding agents closing the loop with issues (finish → update issu | 2026-10-30 | partial |
| MKNKnowledge Graph Memory ServerKnowledge Graph Memory Server | MCPMCP | Model Context Protocol | 让 Agent 跨会话记住用户偏好、项目背景等长期信息。 不适合存放需要审计或追溯的场景（图谱会原地修改）。Scenarios needing cross-session memory that survives restarts and is q | 2026-10-23 | verified |
| MoMOMongoDB MCP ServerMongoDB MCP Server | MCPMCP | MongoDB | 适合：MongoDB 项目的 agent 开发流（查 schema、跑聚合、试写操作）。 不适合：要硬只读沙箱；非 MongoDB 栈。Fits: agent dev flows on the MongoDB stack, Atlas management into agen | 2026-10-30 | partial |
| NeNENeon MCP ServerNeon MCP Server | MCPMCP | Neon | 适合：Neon 栈的 agent 开发流（建分支→试改→合并）、serverless Postgres 管理接 agent。 不适合：非 NFits: agent dev flows on Neon/serverless Postgres, branch-isolated exp | 2026-10-30 | partial |
| NoNONotion MCPNotion MCP | MCPMCP | Notion | 适合：agent 读写团队知识库、把会议记录/任务写进 Notion 数据库、从 wiki 检索上下文。 不适合：要离线/自托管（无本地形态Fits: knowledge-base retrieval into agent context; auto-writing meetin | 2026-10-30 | partial |
| PPLPlaywright MCP ServerPlaywright MCP Server | MCPMCP | Microsoft | 需要浏览器自动化且要保持会话状态的场景： 探索性自动化、自愈测试、长时自主工作流。 不适合高吞吐编码 agent——README 自己建议这Scenarios where an agent must operate a real browser and read the page | 2026-10-23 | verified |
| SeSESentry MCP / ToolkitSentry MCP / Toolkit | MCPMCP | Sentry | 适合：编码 agent 直接查「这个报错在 Sentry 里有没有」、线上事故诊断接 agent 工作流。 不适合：要本地日志分析（不在作用Fits: coding/ops agents plugging Sentry into diagnosis flows; pulling | 2026-10-30 | partial |
| SSESequential Thinking MCP ServerSequential Thinking MCP Server | MCPMCP | Model Context Protocol | 让 Agent 用结构化、可修订、可分支的方式推进复杂问题的分析。 不适合简单任务（增加调用轮次无收益）。Scenarios where a complex problem benefits from being worked through s | 2026-10-23 | verified |
| ShSHShopify Dev MCP ServerShopify Dev MCP Server | MCPMCP | Shopify | 适合：Shopify 开发者给编码 agent 补「平台上下文」——文档检索+schema 探查免模型瞎猜 API 字段； 零鉴权零配置起步Fits: Shopify developers having agents write correct API fields/schema | 2026-10-31 | partial |
| SlSLSlack MCP ServerSlack MCP Server | MCPMCP | Slack（Salesforce） | 适合：编码/研究 agent 取「团队在 Slack 里的决策与上下文」、总结频道、代拟消息。 不适合：要事件触发自动化（MCP 不是触发器Fits: agents grounding answers in real team discussions, channel summa | 2026-10-30 | partial |
| StSTStripe Agent Toolkit / MCPStripe Agent Toolkit / MCP | MCPMCP | Stripe | 适合：Stripe 集成开发时让 agent 查 API 用法、查账户对象、试测试模式操作。 不适合：要「只读保险」硬边界；非 StripeFits: Stripe integration dev flows, rehearsing agent operations in tes | 2026-10-30 | partial |
| SuSUSupabase MCPSupabase MCP | MCPMCP | Supabase | 适合：编码 agent 直接管自家 Supabase 项目（查 schema、跑迁移、看日志）。 不适合：要「只读保险」的硬边界（裁剪能收窄Fits: agent dev flows on the Supabase stack (schema/migrations/logs in | 2026-10-30 | partial |
| TTITime MCP ServerTime MCP Server | MCPMCP | Model Context Protocol | 给 Agent 提供可靠的时间与时区换算能力。 适合需要避免模型猜时间的场景（如日程、跨时区协作）。Scenarios where an agent needs the current date and time to reason abo | 2026-10-23 | verified |
| ZaZAZapier MCPZapier MCP | MCPMCP | Zapier | 适合：一个 server 接多应用（agent 要同时碰 Gmail+Sheets+CRM 时）、按 action 白名单收紧。 不适合：要Fits: agents working across many apps, permission control at action gr | 2026-10-30 | partial |
为什么这里不给总分排名Why there is no overall ranking
这不是省略，是方法上的明确取舍。This is not an omission but a deliberate methodological choice.
- 缺统一实测，就不给分。跨工具的跑分口径不同——题目、预算、执行环境都不一样，A 的 90 分和 B 的 88 分不可比。No shared measurement, no score. Benchmarks across tools use different rules — different questions, budgets and environments. A 90 and a B 88 are not comparable.
- 「未知」是合法答案。查不到就写「未知」并说明为什么，不用推测填充。缺信息本身也是信息。"Unknown" is a legitimate answer. If it cannot be found, we write "unknown" and explain why, instead of filling the gap with guesses. Missing information is itself information.
- 来源链接、类型与核验日都在详情页里，可以逐条回去复核。The source link, its kind and the checked date are all on the detail page, so you can walk any of them back and re-check.
- lifecycle 单独标注。核验日新不等于数据有效——上游归档后数据会失效，所以单独标
active/maintenance/archived。lifecycle is marked separately. A recent check date does not mean the data is still valid — it goes stale when upstream archives the repo, soactive/maintenance/archivedare tracked separately. - 不同族不混排。找现成的、自己搭的、接上与跑起来的是三种角色，坐标系不同；硬合成一张总表只会制造假的可比性。Families are not mixed. Ready-made tools, things you build yourself, and plug-in-and-run components are three different roles with different coordinate systems; forcing them into one master table would manufacture false comparability.
可信度标记怎么读How to read the confidence marks
详情页与上表都带Both the detail pages and the table above carry confidence，它反映, which reflects 本站数据的完整度how complete this site's data is，不是对产品的评价。, not a judgement of the product.
| verified | 11 个维度均有官方源支撑，且核验日在 90 天内all 11 dimensions have official sources and were checked within 90 days |
|---|---|
| partial | 部分维度标为未知，或官方文档不可访问，或核验日超过 90 天some dimensions are marked unknown, or the official docs were unreachable, or checked over 90 days ago |
| stale | 官方已发布重大变化，本站尚未核验the vendor has shipped a significant change that this site has not re-checked |
本站当前的完整度Current completeness on this site
数据层用Recheck this figure with node scripts/audit-gaps.mjs 可随时复核这个数字。.
下表实时计算（本页档案最后核验日 2026-10-31）：Computed from current data; profiles last checked 2026-10-31:
| 大类Category | 已补齐Filled | 说明Meaning |
|---|---|---|
| MCP 工具MCP tools | 328/352 | 官方 README 信息充分，绝大多数维度可直接引述Official READMEs cover most of it, so most dimensions can be quoted directly |
全站合计 1372/1896 维度已补齐Site-wide 1372/1896 dimensions filled（72%）
剩余未补齐项分三类：官方未公开（索引算法、沙箱实现细节本就不对外说明）、需实测才能确定（大仓库表现、CI 无 TTY 行为）、客观渠道不可达。The remaining gaps fall into three kinds: not disclosed by the vendor (indexing algorithms, sandbox internals are simply not documented), requires measurement (behaviour on very large repositories, CI without a TTY), and unreachable through objective channels.
我们选择留白而不是填「已支持」——错误的成本最终由使用者承担。We chose to leave a blank rather than fill in "supported" — the cost of an error is ultimately borne by whoever uses it.
数据来源与复核方式Sources and how to re-check them
全部数据来自开源仓库All data comes from the open-source repository speculcom/ai-handbook（CC BY 4.0）。 (CC BY 4.0).
每个条目都是纯 Markdown + frontmatter，含维度记录、证据链接、核验日与可信度标记。仓库内含：Every entry is plain Markdown plus frontmatter, carrying dimension records and an evidence chain with links, checked dates and confidence marks. The repository contains:
tools/METHODOLOGY.md—— 方法论总纲与核心规则the methodology charter and core rulestools/axes/—— 每个维度的定义与判定标准（含正反例）the definition and decision criteria for each dimension (with positive and negative examples)tools/SCHEMA.md—— 数据格式规范与构建校验规则the data-format spec and build validation rulestools/protocols/<大类><category>/_protocol.md—— 实测协议(the measurement protocol (尚未执行not yet run，故本站暂无实测结论), so this site has no measured conclusions)
发现错误或有新证据，欢迎提 Issue 或 PR——每条修正都会注明依据与影响范围。Found an error or have new evidence? Open an issue or a PR — each correction notes its basis and blast radius.