About MCP_TOOL_POISONING_ATTACKS
MCP_TOOL_POISONING_ATTACKS is a Node.js MCP client and server study project. It runs a configured MCP server and displays its available tools, with Docker and Dev Container setup options.
Key features
- 01MCP client and server example
- 02Loads a configured MCP server
- 03Displays available tools
- 04Docker and Dev Container workflows
Use cases
- 01Study MCP client-server interaction
- 02Inspect tools exposed by a configured server
MCP support
The project includes an MCP client and server. The client reads mcpservers.json and currently uses the first configured server.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEMCP_TOOL_POISONING_ATTACKS
INDEX
ABOUT
MCP Client and MCP Server
ENVIRONMENT
- node.js v22
- mcp sdk
claude-3-5-haiku-20241022
Note
./servers/ipinfo/は現在
npxを使ってMCP Clientから使用する形を取っている。 npxを使ってGitHubのリポジトリからinstallするにはpackage.jsonをリポジトリトップに配置する必要があるため,npx使用のための./package.jsonをリポジトリトップに配置している。
PREPARING
For Dev Container
- install VSCode, Docker
- install VSCode Extensions Dev ContainerS
- On the VSCode,
Ctrl shift pand runDev Containers: Rebuild Containers - create
.envand addANTHROPIC_API_KEY
cat << EOF > mcp_client/.env
ANTHROPIC_API_KEY=your_anthropic_api_key
EOF
- create
mcp_client/mcpservers.json
Docker
- create
.envand addANTHROPIC_API_KEY
cat << EOF > mcp_client/.env
ANTHROPIC_API_KEY=your_anthropic_api_key
EOF
- create
mcp_client/mcpservers.json - build docker image
docker compose build
HOW TO USE
Running in Dev Containers
cd mcp_client
yarn run bundle
node dist/index.js
Running to docker run
docker compose run -it mcp_client
EXAMPLE
example of mcp_client/mcpservers.json
{
"mcpServers": {
"ipinfo": {
"command": "/usr/local/bin/npx",
"args": [
"-y",
"github:RyosukeDTomita/mcp_tool_poisoning_attacks#main",
"ipinfo"
],
"env": {
"PATH": "/usr/local/bin:/usr/bin:/bin"
}
},
"burp": {
"url": "http://localhost:9876/sse"
}
}
}
Note
現状は
mcpservers.jsonの中にある一番上のサーバを使用するようになっている。
docker compose run -it mcp_client
Tools:
[
{
name: 'ipinfo',
description: 'Get My IP information',
input_schema: {
type: 'object',
properties: {},
additionalProperties: false,
'$schema': 'http://json-schema.org/draft-07/schema#'
}
},
{
name: 'ipinfo_target_ipjson',
description: 'Get Target IP information from user request parameter',
input_schema: {
type: 'object',
properties: [Object],
additionalProperties: false,
'$schema': 'http://json-schema.org/draft-07/schema#'
}
}
]
Enter your message: 8.8.8.8の情報を教えて
=====Request to Anthoropic API=====
[ { role: 'user', content: '8.8.8.8の情報を教えて' } ]
=====Response from Anthropic API=====:
{
id: 'msg_012ZGLAfhWKmgDroHELiq6F6',
type: 'message',
role: 'assistant',
model: 'claude-3-5-haiku-20241022',
content: [
{
type: 'text',
text: '8.8.8.8の情報を調べるために、ipinfo_target_ipjsonツールを使用します。'
},
{
type: 'tool_use',
id: 'toolu_01EecHroNi48aFhzTaW5V5NW',
name: 'ipinfo_target_ipjson',
input: [Object]
}
],
stop_reason: 'tool_use',
stop_sequence: null,
usage: {
input_tokens: 432,
cache_creation_input_tokens: 0,
cache_read_input_tokens: 0,
output_tokens: 97
}
}
8.8.8.8の情報を調べるために、ipinfo_target_ipjsonツールを使用します。
=====MCP Server Tool result=====
: {
content: [
{
type: 'text',
text: '{"ip":"8.8.8.8","hostname":"dns.google","city":"Mountain View","region":"California","country":"US","loc":"38.0088,-122.1175","org":"AS15169 Google LLC","postal":"94043","timezone":"America/Los_Angeles","readme":"https://ipinfo.io/missingauth","anycast":true}'
}
]
}
=====Response from Anthropic API after tool use=====
この情報は、IPアドレス8.8.8.8の詳細を示しています。主な特徴は以下の通りです:
1. IP: 8.8.8.8
2. ホスト名: dns.google
3. 所在地:
- 都市: Mountain View
- 地域: カリフォルニア
- 国: アメリカ合衆国(US)
4. 地理的座標: 北緯38.0088、西経-122.1175
5. 組織: AS15169 Google LLC
6. 郵便番号: 94043
7. タイムゾーン: アメリカ/ロサンゼルス
8. エニーキャスト: はい(true)
この8.8.8.8は、Googleが提供する公開DNSサーバーの1つで、一般的に多くのユーザーが利用している信頼性の高いDNSサービスです。
References
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.
