mock-mcp-server

mock-mcp-server

A test MCP server that validates gateway connectivity, identity propagation, and provides mock tools for testing purposes.

Category
访问服务器

README

mock-mcp-server

A test MCP server that runs behind mcp-auth-gateway. It exists to validate the platform plumbing — not to be a real product.

Purpose

This server is used to verify:

  1. gateway → backend MCP server connectivity
  2. path-based /mock/mcp routing
  3. gateway-signed X-MCP-Identity verification
  4. propagation of the user loginid
  5. the Streamable HTTP MCP transport
  6. GitOps / Kubernetes deployment

How it fits in

client ──TLS──> mcp-auth-gateway ──/mock/mcp──> mock-mcp-server (this repo)
                (Keycloak / OAuth,              (verifies X-MCP-Identity,
                 mints X-MCP-Identity)           runs MCP tools)
  • It runs only behind the gateway. External clients reach it exclusively via https://gateway.mcp.aidev.samsungds.net/mock/mcp.
  • It does NOT do Keycloak / OAuth verification itself. All external auth is the gateway's job.
  • It DOES verify the gateway's internal X-MCP-Identity JWT on every /mcp request and rejects anything missing or invalid with 401.
  • Tools read the caller's identity — especially loginid — through a shared get_current_user() helper; they never parse HTTP headers directly.

Identity: X-MCP-Identity

The gateway injects a signed internal JWT:

X-MCP-Identity: <gateway-signed-internal-jwt>

Example payload:

{
  "iss": "mcp-auth-gateway",
  "aud": "mock-mcp-server",
  "sub": "keycloak-user-sub",
  "loginid": "user.loginid",
  "username": "user.name",
  "email": "user@example.com",
  "groups": ["engineering"],
  "scopes": ["mcp:mock:use"],
  "request_id": "01J...",
  "iat": 1730000000,
  "exp": 1730000060
}

Verified on every /mcp request:

  • JWT signature
  • iss == mcp-auth-gateway
  • aud == mock-mcp-server
  • exp (not expired)
  • loginid present
  • sub present

The MVP uses shared-secret HS256 via MCP_IDENTITY_JWT_SECRET. The verifier lives behind a small abstraction (identity.IdentityVerifier) so a production RS256 / JWKS implementation can be dropped in without touching tools or middleware.

loginid usage

loginid is the primary per-user key. Tools use it to build deterministic, user-scoped mock data (mock_profile, mock_search) and to echo the caller's identity (echo, whoami).

Endpoints

Method Path Auth Purpose
GET /healthz none liveness
GET /readyz none readiness
ANY /mcp X-MCP-Identity (JWT) Streamable HTTP MCP endpoint

Requests to /mcp without a valid X-MCP-Identity are rejected at the HTTP layer with 401 Unauthorized before reaching any tool.

MCP tools

Tool Input Returns
whoami – full identity asserted by the gateway
echo message: str { loginid, message }
mock_profile – deterministic mock profile derived from loginid
mock_search query: str deterministic, user-scoped mock search results

Configuration

All configuration is via environment variables.

Variable Default Notes
APP_HOST 0.0.0.0
APP_PORT 8080 container port
MCP_ENDPOINT /mcp Streamable HTTP path
MCP_IDENTITY_ISSUER mcp-auth-gateway required iss
MCP_IDENTITY_AUDIENCE mock-mcp-server required aud
MCP_IDENTITY_ALGORITHM HS256 verifier algorithm
MCP_IDENTITY_JWT_SECRET – required unless dev mode is on
MCP_IDENTITY_DEV_MODE false if true, unauthenticated calls get a dummy user
LOG_LEVEL info
  • MCP_IDENTITY_JWT_SECRET is mandatory in production; the server refuses to start without it.
  • MCP_IDENTITY_DEV_MODE=true is only for local development. It defaults to false and must never be enabled in the cluster.

Local development

Requirements: Python ≥ 3.11 and uv (or pip).

# 1. install
uv venv
uv pip install -e ".[dev]"
# (pip fallback: pip install -r requirements.txt && pip install -e .)

# 2. run the server
export MCP_IDENTITY_JWT_SECRET=dev-secret
export MCP_IDENTITY_DEV_MODE=false
python -m mock_mcp_server.main

Generate a dev identity token

Without the gateway you must supply your own X-MCP-Identity JWT:

export MCP_IDENTITY_JWT_SECRET=dev-secret

python scripts/make-dev-identity-token.py \
  --loginid test.user \
  --subject test-sub \
  --aud mock-mcp-server
# ->  export X_MCP_IDENTITY='<jwt>'

eval "$(python scripts/make-dev-identity-token.py --loginid test.user --subject test-sub)"

Test with curl

/mcp speaks Streamable HTTP (JSON-RPC). In stateless JSON mode you can call a tool directly:

# health (no auth)
curl -s localhost:8080/healthz

# whoami (requires identity)
curl -s localhost:8080/mcp \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "X-MCP-Identity: ${X_MCP_IDENTITY}" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"whoami","arguments":{}}}'

# missing identity -> 401
curl -i -s localhost:8080/mcp \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Test with MCP Inspector

npx @modelcontextprotocol/inspector

In the Inspector UI: choose Streamable HTTP transport, set the URL to http://localhost:8080/mcp, and add a header X-MCP-Identity with your dev token.

Run the tests

uv run pytest        # or: .venv/bin/python -m pytest

Docker

Image name (pushed to cr.aidev.samsungds.net/mcp-platform):

cr.aidev.samsungds.net/mcp-platform/mock-mcp-server:<commit-sha>

Build & push:

COMMIT_SHA=$(git rev-parse --short HEAD)

docker build -t cr.aidev.samsungds.net/mcp-platform/mock-mcp-server:${COMMIT_SHA} .
docker push cr.aidev.samsungds.net/mcp-platform/mock-mcp-server:${COMMIT_SHA}

The image runs as a non-root user, exposes 8080, and ships a HEALTHCHECK against /healthz. Tag images by commit SHA (no latest in the cluster) so GitOps pins an exact, reproducible build.

Deployment (mcp-platform-gitops)

Full Kubernetes / GitOps manifests live in mcp-platform-gitops, not here. This repo only ships the app and image. The gitops repo should deploy it as a ClusterIP Service with these values:

Setting Value
Namespace mcp-gateway
Service name mcp-mock
Service type ClusterIP
Container port 8080
Internal DNS http://mcp-mock.mcp-gateway.svc.cluster.local:8080
Internal MCP URL http://mcp-mock.mcp-gateway.svc.cluster.local:8080/mcp
External MCP URL https://gateway.mcp.aidev.samsungds.net/mock/mcp (gateway)

Required Secret (mcp-identity-jwt, key secret) holds the shared HS256 secret, which must match the gateway's signing secret.

Container env:

env:
  - name: MCP_IDENTITY_JWT_SECRET
    valueFrom:
      secretKeyRef:
        name: mcp-identity-jwt
        key: secret
  - name: MCP_IDENTITY_ISSUER
    value: mcp-auth-gateway
  - name: MCP_IDENTITY_AUDIENCE
    value: mock-mcp-server

Probes: GET /healthz (liveness), GET /readyz (readiness), both on 8080.

Do NOT create a public Ingress

This service must never be exposed with its own public Ingress / LoadBalancer.

  • It performs no external authentication (no Keycloak, no OAuth). It only trusts the gateway-signed X-MCP-Identity. A direct public route would let anyone reach /mcp while bypassing the gateway's real authentication.
  • All external access must go through the gateway, which authenticates the user, mints X-MCP-Identity, and routes /mock/mcp to this ClusterIP Service.

The only external entry point is:

https://gateway.mcp.aidev.samsungds.net/mock/mcp

Project layout

src/mock_mcp_server/
  __init__.py
  main.py          # ASGI app factory + uvicorn entrypoint
  settings.py      # env-based configuration (fails fast)
  identity.py      # X-MCP-Identity JWT verification (HS256; RS256-ready)
  middleware.py    # pure-ASGI identity enforcement (401) + scope injection
  context.py       # McpUserContext + get_current_user()
  tools.py         # whoami / echo / mock_profile / mock_search
scripts/
  make-dev-identity-token.py
tests/

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选