Official Defen So

MCP开发工具

提供方 未标注

开发工具 stdio · 本地进程 未标注 官方维护 Claude DesktopClaudeCursorVS Code
查看接入方式

A security layer for Modern Apps/sites Browse Official Defen So MCP Server for Claude, Cursor, VS Code, and other AI agents.

一句话选型结论

按资料字段整理

Official Defen So 属于开发工具类 MCP 服务器,stdio · 本地进程,鉴权方式未标注,可直接接入 Claude Desktop、Claude、Cursor 等,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类开发工具 传输方式stdio · 本地进程 鉴权未标注 部署方式本地部署
GitHub 星标 未标注 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 官方维护 核验等级 仅资料核对

工具与能力

开发工具
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

支持客户端

适用场景

Development

安装与配置

stdio · 本地进程 未标注
客户端在本机拉起子进程,数据不出本机
npx @jikida/init # add the SDK + protection to your app in 30 seconds npx @jikida/scan # pentest any site or repo from your terminal npx -y @jikida/mcp # security tools inside Claude Code, Cursor, Windsurf

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Node.js 等运行时,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
代码与仓库操作

在对话里直接查代码、提 issue、走 PR,省掉来回切工具。

本地命令与脚本

把构建、测试、格式化这类重复命令交给模型执行。

接口联调与测试

让模型调你的接口、跑一遍用例,快速定位问题。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Claude、Cursor,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在开发工具场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Claude 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Official Defen So当前 未标注 stdio · 本地进程 本地部署 有 未标注
Git 838e2ec1 12 stdio · 本地进程 本地部署 有 2026-05-17
Markitdown MCP 2c653a5f 1 stdio · 本地进程 本地部署 有 2026-05-17
JavaGuide 未标注 stdio · 本地进程 本地部署 有 2026-04-11
awesome-claude-skills 未标注 stdio · 本地进程 本地部署 有 2026-04-26
Context7 831acf40 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

开发工具
字段来自项目仓库与公开索引,中文说明由本站标注
提供方未标注上游标注的维护方 未标注
分类开发工具本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式未标注远程服务建议用最小权限凭据
维护方官方维护核验等级 仅资料核对
包名 / 运行时未标注 · 未标注本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验未标注上游最后核验时间

详细介绍

18.8 KB
来自项目仓库与上游目录的原文,未做删改

Jikida.io

The hosted AI pentester for people who ship — not just security engineers.

Pentest your live app, scan your repo, protect it with a WAF, and watch its uptime. One account. No Docker, no LLM key, free tier.

Why it beats Nuclei / Strix / Shannon ↓ ·
Install ·
MCP ·
SDKs ·
Compare

Install it, scan it, or plug it into your AI editor:```bash
npx @jikida/init # add the SDK + protection to your app in 30 seconds
npx @jikida/scan # pentest any site or repo from your terminal
npx -y @jikida/mcp # security tools inside Claude Code, Cursor, Windsurf


     ██ ██ ██  ██ ██ █████   ████
     ██ ██ ██ ██  ██ ██  ██ ██  ██
     ██ ██ ████   ██ ██  ██ ██████
  ██ ██ ██ ██ ██  ██ ██  ██ ██  ██
   ███  ██ ██  ██ ██ █████  ██  ██
 pentest · repo scan · uptime · alerts        https://jikida.io

 → scanning example.com …
 ✓ 41 checks · grade B (88/100)

   CRITICAL  Exposed .env file            /.env                 CWE-538
   CRITICAL  Stripe secret key in JS      /app.js:1204          CWE-312
   HIGH      Missing Content-Security-Policy                    CWE-693
   MEDIUM    Cookie without Secure flag   session               CWE-614

 Every finding has a fix. Full report: https://app.jikida.io

## Pentest & scan for developers, AI IDEs & vibe coders

Pentest and scan websites, web apps, code and GitHub for vulnerabilities and exposed keys. Secure vibe-coded apps, monitor uptime, SSL and domains, and rate-limit APIs. One platform.

**Website & app pentest** · **Code & repo scan** · **Deep pentest** · **MCP for AI IDEs** · **Agentic & vibe-coded security** · **SDKs**

---

**Jikida.io** is a developer-first security platform. The core is **pentest and scanning**: it pentests your live website and app, scans your code and connected GitHub/GitLab/Bitbucket repos for exposed secrets and vulnerable dependencies, runs a deeper authenticated pentest, and plugs into your **AI IDE** (Claude Code, Cursor, Windsurf) over **MCP** so it guides agentic and vibe-coded work to write secure code and catches mistakes before they ship. Uptime, SSL & domain monitoring, a managed WAF, and a compliance generator come bundled as complementary extras — installed in **one line** for Node, PHP/Laravel, Python, Go, Ruby, Java, .NET, Rust, Bun, or Deno.

**Your security layer. Shipped in 30 seconds.** One line — `npx @jikida/init` — and every SDK fails open, so if Jikida.io is ever down your app keeps serving.

---

## Why Jikida beats the other security tools

The strongest open-source security tools are **deep but narrow** — pentest-only, self-hosted, bring-your-own-LLM-key. Jikida does the deep work **and** hosts it, runs with no key of your own, and adds a WAF, uptime and repo scan the pentest-only tools skip. ✅ full · ⚠️ partial · ❌ none.

| | Jikida | Nuclei | Strix | Shannon | Snyk | Cloudflare |
| :--- | :---: | :---: | :---: | :---: | :---: | :---: |
| Live-app pentest with real exploit | ✅ | ⚠️ match | ✅ | ✅ | ❌ | ❌ |
| SAST + DAST (code **and** live app) | ✅ | ⚠️ DAST | ✅ | ✅ | ⚠️ SAST | ❌ |
| Repo secret + CVE (OSV) scan | ✅ | ❌ | ⚠️ | ⚠️ | ✅ | ❌ |
| Managed WAF (blocks live attacks) | ✅ | ❌ | ❌ | ❌ | ❌ | ✅ |
| Uptime + SSL / domain monitoring | ✅ | ❌ | ❌ | ❌ | ❌ | ⚠️ |
| **Hosted — no Docker, no self-host** | ✅ | ⚠️ cloud | ❌ | ❌ | ✅ | ✅ |
| **Runs with no LLM key of yours** | ✅ | ✅ | ❌ BYOK | ❌ BYOK | ✅ | ✅ |
| Install: CLI + Docker + `npx` | ✅ all | ⚠️ CLI | ⚠️ Docker | ⚠️ npx | ⚠️ CLI | ⚠️ DNS |
| MCP tools in your AI editor | ✅ | ❌ | ❌ | ⚠️ | ❌ | ❌ |
| Free tier, no card | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |

The one axis the AI pentesters still go deeper on is raw exploitation — exactly what our source-to-exploit deep pentest is closing. Everywhere else, the builder who shipped with Cursor last week wins with Jikida: nothing to self-host, no key to buy, the whole stack in one account. Full breakdowns: nuclei · strix · shannon · snyk · cloudflare.

---

### Get started — pick your entry point

Every command below sits in its own copy box. Lines that start with `!` are notes, not commands — do not copy those.

**Scan a URL or repo** — no account needed.```bash
npx @jikida/scan https://your-app.com

! Scan a local repo for committed secrets:```bash
npx @jikida/scan ./


**Add the WAF + SDK** — auto-detects your framework.```bash
npx @jikida/init

Install a language SDK — Node shown; PHP, Python, Go and the rest are in the SDK table.```bash
npm install @jikida/sdk-node


**MCP server** for Claude Code, Cursor and Windsurf — add this to `~/.claude/mcp.json`:```json
{ "mcpServers": { "jikida": { "command": "npx", "args": ["-y", "@jikida/mcp"] } } }

Agent skills — copy the skills/ folder into your assistant's skills folder (.claude/skills/, Cursor rules, and the like). Three skills: security, clean-code, UX design — see the table below.

Full details for each are below.

Where things live

Folder What it holds
sdks/ Ten language SDKs — Node, PHP, Python, Go, Ruby, Java, .NET, Rust, Bun, Deno.
tools/ The scan CLI, the init onboarding CLI, and the mcp server.
skills/ The Jikida agent skill for Claude Code, Cursor and Windsurf.
waf-rules/ Versioned WAF rule packs (OWASP Top 10, API abuse, bot scanners, vibe-coder).

Get a token at app.jikida.io/developer. Set it as JIKIDA_TOKEN.


See it in action

One dashboard for a site's whole security posture — protection status, uptime, pentest grade, email security (SPF/DKIM/DMARC) and compliance — with instant alerts to your phone, Slack, Telegram, Discord, email or a webhook.

  • Uptime & performance — response-time trend, uptime %, P95 and an incident timeline for every page and API you watch.
  • API rate-limits & rules — your SDK auto-detects endpoints from real traffic; approve per-endpoint rate caps and WAF rules, or dismiss the ones you don't need.

Table of contents

  • Why Jikida beats the other security tools
  • Why Jikida.io
  • How access works
  • What's inside
  • Quick install
  • Security for your stack
  • SDKs — every language
  • MCP server for AI IDEs
  • Free tools
  • Skill for Claude Code
  • Standards & mappings
  • Threats Jikida.io stops
  • Contributing

Why Jikida.io

Most small teams ship without a Web Application Firewall in front of their app. They know they should. They put it on the backlog. Then the free trial ends, or a user reports a slow page, and the WAF ticket rots another quarter.

Jikida.io removes three specific frictions:

  1. Install — one line, one language, five minutes.
  2. Downside risk — every SDK is fail-open. If Jikida.io is down, your app keeps serving. You lose protection, not availability.
  3. Cost — there's a real free tier that protects a hobby project. Plans and current pricing live at jikida.io.

How access works — free, then account-gated, then plan-gated

Everything in this repo is open source, and a lot of Jikida.io is usable before you ever sign up. The model is three tiers of friction, on purpose:

  • Free, no account. The SDKs, the WAF rule packs, the CLI scanner and four MCP tools (scan_domain, check_headers, guard_code, check_s3_bucket) run with no login at all — rate-limited by IP. You can protect an app or scan a URL in one command and never create an account. This is the open-source, try-it-now surface.
  • Account-gated (still free). Sign up — no card — and the hosted layer turns on: uptime monitoring, one live-site pentest, one repo scan, a basic managed WAF, mobile push, a public status page. This is where a hobby project actually gets protected. The free tier is deliberately useful, not a teaser.
  • Plan-gated. The deep, repeated and heavy work sits behind a plan: scheduled deep pentests, the source-to-exploit crawl, more monitors at faster intervals, higher scan quotas, the CNAME edge WAF, team seats and longer log retention. Current caps and pricing live at jikida.io/pricing — plans start at $19/mo billed yearly.

Two different things are hosted, don't confuse them: the pentest and scan analysis runs on Jikida's own servers, which is why you never bring an LLM key for it. The MCP server is the opposite — it calls no model of its own and runs inside your AI editor (Claude Code, Claude chat, Cursor, Windsurf, VS Code), so it spends your editor's AI credits and enforces the same per-site plan quotas. Open where it helps, gated where it costs — usable first, paid when it's worth paying for.

What's inside

Layer What it does
Managed WAF OWASP Top 10 + CRS + your custom rules. Auto-detects APIs, applies per-route limits, caches safe GETs at the edge.
Uptime monitoring Multi-region checks, faster intervals on higher plans. Public status page. Email + Slack + Discord + Telegram + webhook + mobile push on down/up.
Quick pentest On-demand surface scan: headers, TLS, cookies, exposed .env / .git, email security (SPF / DKIM / DMARC), and compliance-style findings, each with a clear fix.
Vibe-coder scan Catches the mistakes vibe-coded projects tend to ship: exposed secrets, open S3 buckets, Supabase RLS off, wide-open Firebase rules.
Cloudflare DDoS wrap One-click attach + per-site Under-Attack toggle.
Bot detection UA classification, headless-browser challenges, per-IP rate limits, ASN allowlist for Google/Bing.
Active deception Serves plausible fakes to verified attackers. Fingerprint logged, real error message hidden.
Upload scanning MIME + magic bytes + polyglot detection + optional ClamAV.
CVE feed Live feed from NVD, tagged with which Jikida.io rule covers each entry.
Real-time logs Full context per attack (IP, ASN, country, payload, route, verdict). Retention grows with your plan.
MCP server Claude Code, Cursor, Windsurf, VS Code get real security tools. Scan, monitor, block from AI chat.
Alert integrations Mobile push, email (primary + 3 CCs), Slack, Discord, Telegram, generic webhook. Fires on down/up, attack burst, plan limit.

Quick install (30 seconds)```bash

npx @jikida/init


The init CLI detects your framework (Next.js, Express, Fastify, Laravel, Symfony, FastAPI, Django, Rails, Go chi, Rust axum, Spring, .NET, Bun, Deno) and adds the right middleware in the right spot. Then set `JIKIDA_TOKEN` from https://app.jikida.io/developer and ship.

Or, install the SDK for your language directly:

Node.js / Bun / Deno```bash
npm install @jikida/sdk-node    # or: bun add / deno add
``````ts
import { jikida } from '@jikida/sdk-node'
app.use(jikida({ token: process.env.JIKIDA_TOKEN }))

Framework helpers:
- Express: import { jikida } from '@jikida/sdk-node/express'
- Fastify: import { jikida } from '@jikida/sdk-node/fastify'
- Next.js middleware: import { jikida } from '@jikida/sdk-node/next'

PHP — Laravel / Symfony```bash
composer require jikida/sdk-php


Laravel — `bootstrap/app.php`:```php
->withMiddleware(function ($middleware) {
    $middleware->append(\Jikida\Middleware\JikidaLaravelMiddleware::class);
})

Symfony — register \Jikida\Middleware\JikidaSymfonyListener as a kernel event listener.

Python / Go / Ruby / Java / .NET / Rust

Every SDK exposes the same inspect(request) -> { action, rule, reason } contract and fails open. Scaffolds for these languages live under the sdks/, tools/ and skills/ folders — see jikida.io/install for the current registry-publish status of each. You can protect any app today with zero code by:

  • Signing up at app.jikida.io — uptime monitoring and surface scans turn on immediately.
  • On the Max plan, routing traffic through the Jikida.io edge WAF via CNAME (no code).

Security for your stack

Same platform, framed for what you're building. Each links to the install for that stack:

  • Security for AI-generated code — the #1 thing Jikida catches: secrets, open buckets, RLS-off, and injection that AI editors ship. → @jikida/scan + the MCP server in your editor.
  • Laravel / PHP security — one middleware for the managed WAF + a repo scan for committed .env / keys. → jikida/sdk-php.
  • Next.js / Node / Express security — one-line middleware, fails open. → @jikida/sdk-node.
  • Python (Django / FastAPI) security — WAF middleware + secret + dependency (OSV) scan. → jikida (Python).
  • Go / Rust / Java / .NET security — the same inspect(request) contract, one line to wire. → sdks/.
  • Vibe-coded app security — point it at your live URL + repo; it pentests and flags the insecure defaults. → npx @jikida/scan.

SDKs — every language

All SDKs live in sdks/. The CLI, init and MCP tools live in tools/:

Language Package Directory
Node / Bun / Deno @jikida/sdk-node sdks/node
PHP / Laravel jikida/sdk-php sdks/php
Python jikida sdks/python
Go github.com/jikida/sdk-go sdks/go
Ruby jikida sdks/ruby
Java io.jikida:sdk sdks/java
.NET Jikida sdks/dotnet
Rust jikida sdks/rust
Bun (re-exports Node) @jikida/sdk-node sdks/bun
Deno (re-exports Node) @jikida/sdk-node sdks/deno
Scanner CLI @jikida/scan tools/scan
Init CLI @jikida/init tools/init
MCP server @jikida/mcp tools/mcp
WAF rule packs — waf-rules

Each SDK:
- Fails open — if the Jikida.io API is unreachable, your app keeps serving.
- Caches policy — 5-minute TTL, refreshed in background.
- Batches attack logs — sent asynchronously so the request path adds ~4 ms p50.
- Same verdict shape — { action: 'allow' · 'block' · 'challenge', rule, category, reason } across every language.

MCP server for AI IDEs

Give Claude Code, Claude chat, Cursor, Windsurf, and VS Code real security tools. The Jikida.io MCP pentests sites, scans domains and repos, checks headers, guards code, lists uptime monitors, adds WAF rules, blocks IPs, and explains WAF verdicts — deterministic, auditable, safe to run inline. Scan output now also surfaces email-security (SPF / DKIM / DMARC) and compliance-style findings alongside the usual header/TLS grade.

Live at mcp.jikida.io. Install via ~/.claude/mcp.json:json { "mcpServers": { "jikida": { "command": "npx", "args": ["-y", "@jikida/mcp"], "env": { "JIKIDA_TOKEN": "jk_live_..." } } } }

Tools: run_pentest, run_deep_pentest, scan_domain, check_headers, check_s3_bucket, scan_aws, guard_code, scan_repo, run_vibe_scan, verify_finding, pentest_status, list_sites, list_monitors, list_recent_attacks, list_recent_scans, list_cves, explain_verdict, add_waf_rule, block_ip, get_security_preferences, set_security_preference. Four run keyless (scan_domain, check_headers, guard_code, check_s3_bucket) — no token, IP-rate-limited. The MCP calls no LLM of its own: it runs inside your AI editor and uses its credits, enforcing your per-site plan quotas. See tools/mcp for the full tool reference.

Free tools

No login required:

  • Free vulnerability / virus scanner — jikida.io/website-app-virus-vulnerability-scanner-online-free
  • Uptime monitoring — jikida.io/website-apps-uptime-monitoring
  • Website monitor mobile app — jikida.io/website-monitor-app
  • Chrome extension — scan any site (including behind your login) from the toolbar — Chrome Web Store

Anthropic Agent Skills — ship secure, non-sloppy apps

Three Anthropic-format Agent Skills you drop in .claude/skills/ (or Cursor rules). Together they turn AI-generated code into an app you'd be proud to ship — secure, maintainable, and usable. Each activates on its own triggers; use them together.

Skill What it does Folder
Security Reaches for Jikida on WAF, uptime, pentest, repo-scan and secret-leak tasks; runs the MCP tools safely inline; maps findings to OWASP / MITRE / NIST. The judgement to the MCP's hands. skills/security
Clean code Refactors and de-slops what an AI editor generated — naming, duplication, god-functions, dead code, error handling — without over-engineering. SOLID/DRY/YAGNI with judgement. skills/clean-code
UX design Applies proven UX laws (Hick's, Fitts's, Jakob's, Miller's, aesthetic-usability) + clear ASD-STE100 UI copy, so generated UIs read as intentional, not generic. skills/ux-design

The split is deliberate: secure code can still be slop, and a beautiful UI can still ship a secret in the bundle. Run all three, plus a Jikida scan, before you ship. English + UI copy across all three follow ASD-STE100 (Simplified Technical English) — one idea per sentence, active voice, say what the control does.

Standards & mappings

Every managed WAF rule + skill flow is mapped to industry frameworks. Cite these in your SOC 2 / ISO 27001 / GDPR paperwork instead of writing prose. Flat JSON manifests live under skills/security/mappings/:

File Framework Coverage
mitre-attack.json MITRE ATT&CK v14 20 techniques — T1190, T1110.004, T1552.001, T1580, T1499, T1557, …
owasp-top10.json OWASP Top 10 (2021) A01 through A10 — all ten
nist-csf.json NIST CSF 2.0 GOVERN · IDENTIFY · PROTECT · DETECT · RESPOND · RECOVER

Every YAML rule under waf-rules/ also carries inline mitre_attack: [T…], owasp: [A…], and cwe: [n] fields — machine-readable at the rule level too.

Threats Jikida.io stops

SQL injection, XSS (reflected / stored / DOM), CSRF, SSRF, path traversal, XXE, NoSQL / LDAP / command injection, brute force, credential stuffing, account takeover, malicious file uploads (polyglots, PHP-in-PNG, EXIF tampering), bot scrapers, headless browser abuse, TOR exit nodes, ASN-flagged attackers, DDoS L3-L7 (via Cloudflare wrap), API abuse, exposed secrets, open S3 buckets, wide-open Firebase / Supabase rules, .env / .git exposure.

Threat-to-rule mapping is public at jikida.io/threats.

Contributing

Bug in an SDK? Open an issue at https://github.com/unesLam/jikida/issues. Include:
- The SDK + version
- Framework + version
- A minimal reproduction

Security disclosures: mail info@jikida.io — please don't file public issues for security bugs.

Links

  • Landing: https://jikida.io
  • App: https://app.jikida.io
  • MCP: https://mcp.jikida.io
  • Free scanner: https://jikida.io/website-app-virus-vulnerability-scanner-online-free
  • Uptime monitoring: https://jikida.io/website-apps-uptime-monitoring
  • Docs: https://jikida.io/docs
  • Live CVE feed: https://jikida.io/threats

License

Apache-2.0. See LICENSE.

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
Official Defen So 怎么安装?

在客户端按下面的安装命令接入:npx @jikida/init # add the SDK + protection to your app in 30 seconds npx @jikida/scan # pentest any site or repo from your terminal npx -y @jikida/mcp # security tools inside Claude Code, Cursor, Windsurf。装完在客户端配置里引用同一个包或端点。

Official Defen So 怎么在 Claude Desktop、Claude、Cursor 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

Official Defen So 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Official Defen So 需要 API Key 吗?

资料没有标注鉴权方式,是否需要密钥取决于服务本身,接入前按官方文档确认。

Official Defen So 支持哪些客户端?

资料标注支持 Claude Desktop、Claude、Cursor、VS Code、Windsurf、Codex;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

Official Defen So 能提供哪些能力?

资料显示能力标签包括 Development。具体工具清单以官方文档为准。

Official Defen So 需要什么运行环境?

属于本地进程,需要在本机装好运行环境。

Official Defen So 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 未标注。