/ Каталог / Песочница / inspector
← Все серверы ↗ GitHub
● Сообщество MCPJam ⚡ Сразу

inspector

автор MCPJam · MCPJam/inspector

The debugger you wish you had when building MCP servers — JSON-RPC tracing, tool evals across LLMs, OAuth conformance, CI integration.

MCPJam Inspector is a dev platform for MCP servers. Web app, desktop app, terminal, and Docker. Debug tool calls, chat with any LLM against your server, run eval tests across models, and integrate with CI to catch regressions on PR.

▶ Смотреть демо 📦 Установить 💡 Сценарии

Зачем использовать

Ключевые функции

Живое демо

Как выглядит на практике

inspector.replay ▶ готово
0/0

Установка

Выберите клиент

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "inspector": {
      "command": "npx",
      "args": [
        "-y",
        "inspector"
      ],
      "_inferred": true
    }
  }
}

Откройте Claude Desktop → Settings → Developer → Edit Config. Перезапустите после сохранения.

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "inspector": {
      "command": "npx",
      "args": [
        "-y",
        "inspector"
      ],
      "_inferred": true
    }
  }
}

Cursor использует ту же схему mcpServers, что и Claude Desktop. Конфиг проекта приоритетнее глобального.

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "inspector": {
      "command": "npx",
      "args": [
        "-y",
        "inspector"
      ],
      "_inferred": true
    }
  }
}

Щёлкните значок MCP Servers на боковой панели Cline, затем "Edit Configuration".

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "inspector": {
      "command": "npx",
      "args": [
        "-y",
        "inspector"
      ],
      "_inferred": true
    }
  }
}

Тот же формат, что и Claude Desktop. Перезапустите Windsurf для применения.

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "inspector",
      "command": "npx",
      "args": [
        "-y",
        "inspector"
      ]
    }
  ]
}

Continue использует массив объектов серверов, а не map.

~/.config/zed/settings.json
{
  "context_servers": {
    "inspector": {
      "command": {
        "path": "npx",
        "args": [
          "-y",
          "inspector"
        ]
      }
    }
  }
}

Добавьте в context_servers. Zed перезагружается автоматически.

claude mcp add inspector -- npx -y inspector

Однострочная команда. Проверить: claude mcp list. Удалить: claude mcp remove.

Сценарии использования

Реальные сценарии: inspector

How to debug why your MCP tool isn't being called

👤 MCP server developers ⏱ ~15 min intermediate

Когда использовать: You wrote a tool, Claude doesn't use it, and you don't know why.

Предварительные требования
  • Node — nodejs.org or nvm
Поток
  1. Launch Inspector
    npx @mcpjam/inspector@latest and open the printed URL.✓ Скопировано
    → UI opens in browser
  2. Connect your server
    In the UI, add your server (stdio or HTTP) and connect.✓ Скопировано
    → Tools list populates
  3. Simulate a prompt
    In the Chat tab, ask a question that should trigger your tool. Watch the JSON-RPC trace.✓ Скопировано
    → Either tool fires (good) or you see why the model rejected it (bad description, missing input schema)

Итог: Concrete reason (wrong tool description, schema issue, model choice) with a visible fix.

Подводные камни
  • Testing with a weaker model that never uses tools — Test across multiple models in the Evals tab — Haiku, Sonnet, GPT-4o — behavior varies

How to run eval tests on PRs for your MCP server

👤 Teams maintaining production MCPs ⏱ ~30 min advanced

Когда использовать: You want to catch 'oops that prompt no longer picks my tool' in CI.

Поток
  1. Author evals
    In Inspector, create eval cases: (prompt, expected_tool_called, expected_args).✓ Скопировано
    → Eval suite saved
  2. Wire into CI
    Use the CLI to run evals on every PR; fail the build on regression.✓ Скопировано
    → .github/workflows/mcp-eval.yml runs evals

Итог: PRs that break tool selection don't merge.

Подводные камни
  • Evals are flaky when temperature > 0 — Use temperature 0 in eval runs
Сочетать с: github

How to diagnose a broken OAuth flow on your MCP server

👤 MCP server devs shipping OAuth-protected tools ⏱ ~20 min advanced

Когда использовать: Tokens aren't refreshing, or redirect returns 400.

Поток
  1. Enable OAuth debug in Inspector
    Configure your server's OAuth endpoints and click 'Test flow'.✓ Скопировано
    → Step-by-step trace of authorize, token, refresh
  2. Identify the break
    Inspector highlights nonconforming steps (wrong content-type, missing PKCE, bad scope).✓ Скопировано
    → Specific failure node

Итог: Conforming OAuth implementation.

Комбинации

Сочетайте с другими MCP — эффект x10

inspector + github

Run Inspector evals in a GitHub Action on PRs for your MCP server

Set up a GitHub Action using @mcpjam/inspector CLI to run evals on every PR touching src/tools/.✓ Скопировано

Инструменты

Что предоставляет этот MCP

ИнструментВходные данныеКогда вызыватьСтоимость
(inspector app) connect-server transport: stdio|http, command or url First step free
(inspector app) execute-tool tool, args Manual testing depends on tool
(inspector app) chat-and-inspect model, prompt End-to-end simulation 1 LLM call
(inspector app) run-evals suite_id CI regression testing N LLM calls

Стоимость и лимиты

Во что обходится

Квота API
Your LLM/eval keys determine the budget
Токенов на вызов
Evals can burn tokens — limit concurrency + size
Деньги
Inspector itself is free. Hosted app at app.mcpjam.com has tiers.
Совет
Run evals on Haiku for cheap smoke tests; reserve Sonnet/Opus for nightly runs.

Безопасность

Права, секреты, радиус поражения

Минимальные скоупы: Whatever scopes your MCP server needs LLM provider keys for evals
Хранение учётных данных: Local env for CLI; workspace secrets for the hosted app
Исходящий трафик: Tool calls stay between your MCP and LLM; hosted app stores eval results in your workspace
Никогда не давайте: Don't upload production OAuth client secrets to the hosted app without review

Устранение неполадок

Частые ошибки и исправления

Can't connect to my stdio server

Make sure the command Inspector runs matches what Claude Desktop runs — working dir and env vars included.

Chat tab says 'model not configured'

Add your Anthropic/OpenAI key in Settings.

Evals pass locally, fail in CI

Env mismatch — pin the same model + temperature in CI as locally.

OAuth test reports 'state mismatch'

Your server isn't preserving the state param; compare to the spec in Inspector's conformance report.

Альтернативы

inspector в сравнении

АльтернативаКогда использоватьКомпромисс
@modelcontextprotocol/inspector (official)You want the first-party reference inspectorSmaller feature surface; no evals/CI
mcp-client-for-ollamaYou want a terminal-only Ollama-backed clientNot a debugger, just a client

Ещё

Ресурсы

📖 Читать официальный README на GitHub

🐙 Открытые задачи

🔍 Все 400+ MCP-серверов и Skills