A remote Model Context Protocol (MCP) server that exposes hardcoded application security testing data (applications, scans, and issues) so that LLM-based clients can query security posture using natural language.
All server responses are hardcoded mock data intended for demonstration and testing purposes.
The server exposes the following MCP tools:
| Tool | Description |
|---|---|
get_applications |
List all onboarded applications. |
get_application_details |
Get details for a single application by ID. |
get_scans |
List scans, optionally filtered by application. |
get_scan_details |
Get details for a single scan execution. |
get_issues |
Query issues by application, scan, severity, or status. |
get_issue_details |
Get full details for a specific issue (file, trace, remediation). |
get_dashboard_summary |
Aggregate posture summary across all apps. |
It also exposes a prompt named sentinelscan_doc that loads usage rules,
ID conventions, and allowed enum values into the model's context window.
The server runs over Streamable HTTP on the /mcp endpoint, which is the
standard transport for remote MCP servers. By default it listens on
0.0.0.0:8000.
Every request to /mcp must include both of the following Key ID / Key
Secret headers:
| Header | Value |
|---|---|
X-Key-Id |
sentinel-demo-key-id-12345 |
X-Key-Secret |
sentinel-demo-key-secret-abcdef67890 |
These credentials are hardcoded in server.py as the KEY_ID and
KEY_SECRET constants. There is no real authentication logic - a Starlette
middleware simply compares the incoming headers against those constants and
returns HTTP 401 Unauthorized on mismatch.
Example check with curl:
curl -i https://<your-project>.vercel.app/mcp \
-H "X-Key-Id: sentinel-demo-key-id-12345" \
-H "X-Key-Secret: sentinel-demo-key-secret-abcdef67890"Omitting or changing either header returns:
{
"error": "Unauthorized",
"message": "Missing or invalid credentials. Provide the 'X-Key-Id' and 'X-Key-Secret' headers on every request to the SentinelScan Cloud MCP Server."
}pip install -r requirements.txt
python server.pyThe MCP endpoint will be available at:
http://localhost:8000/mcp
docker build -t sentinelscan-cloud-mcp .
docker run -p 8000:8000 sentinelscan-cloud-mcpThe repo is pre-configured to deploy as a Vercel Python serverless function so anyone can use the server remotely over the public internet.
Files involved:
api/index.py— Vercel entrypoint. Imports the FastMCP instance fromserver.pyand exposes its Streamable-HTTP Starlette app asapp, which Vercel's Python runtime serves automatically.vercel.json— rewrites/mcp(and/mcp/*) to/api/indexso clients can use the canonical MCP path.requirements.txt— picked up by Vercel to install themcpSDK.server.py— runs FastMCP instateless_http=Truemode, which is required on serverless platforms since each request is handled by a fresh function invocation and no session state can be preserved between calls.
Deploy via the Vercel CLI:
npm i -g vercel
vercel login
vercel # preview deployment
vercel --prod # production deploymentOr deploy from GitHub:
- Push this repo to GitHub (already done on branch
claude/create-remote-mcp-server-PMsnL). - Go to https://vercel.com/new and import the repository.
- Framework preset: Other. Leave build/install commands empty —
Vercel will auto-detect
requirements.txtand theapi/directory. - Click Deploy.
After deployment your MCP endpoint will be:
https://<your-project>.vercel.app/mcp
Note: On Vercel's Hobby plan, serverless functions have a 10 second execution timeout. That's plenty for this server since all responses are hardcoded, but if you later wire it up to slow upstream APIs you may need the Pro plan (60s) or longer.
Point any MCP-compatible client (Claude Desktop, Claude Code, or a custom agent built on the Anthropic SDK) at the server URL:
{
"mcpServers": {
"sentinelscan-cloud": {
"url": "http://localhost:8000/mcp",
"transport": "http",
"headers": {
"X-Key-Id": "sentinel-demo-key-id-12345",
"X-Key-Secret": "sentinel-demo-key-secret-abcdef67890"
}
}
}
}For a Vercel-hosted deployment, replace the URL:
{
"mcpServers": {
"sentinelscan-cloud": {
"url": "https://<your-project>.vercel.app/mcp",
"transport": "http",
"headers": {
"X-Key-Id": "sentinel-demo-key-id-12345",
"X-Key-Secret": "sentinel-demo-key-secret-abcdef67890"
}
}
}
}All applications, scans, and issues returned by the server are hardcoded in
server.py. There is no external database or network call - the server
is entirely self-contained and safe to run anywhere for demos and testing.