mcp-server-starter-sample
Provides a tool for executing a single parameterized SELECT query against a local SQLite example database and returning the matching rows.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-server-starter-sample Run a parameterized SELECT for note id 1 and return its title and body."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mcp-server-starter-sample
Free sample: two local MCP tools you can run on your own machine.
SQLite query — a single parameterized
SELECTagainst a tiny example databaseHTTPS fetch — request a host you have already put on an allowlist
This is a November 2026 snapshot, built with the official TypeScript MCP SDK v2 (@modelcontextprotocol/server 2.4.0, specification 2026-07-28). It is not an official Anthropic, Cursor, or Windsurf product. It is not a full production server, and it does not include a general-purpose shell.
Requirements
Node.js 20 or later (Node 22+ recommended for the built-in
node:sqlitemodule)
Related MCP server: Catalog Sentinel
Quick start
npm install
npx tsx src/index.tsThe process waits on stdin for an MCP client. To try the SQLite tool you can also inspect the sample database at data/example.sqlite.
Cursor
Add this to .cursor/mcp.json or ~/.cursor/mcp.json (Cursor MCP docs):
{
"mcpServers": {
"mcp-server-starter-sample": {
"command": "npx",
"args": ["tsx", "ABS/PATH/TO/mcp-server-starter-sample/src/index.ts"],
"env": {
"MCP_FETCH_ALLOWLIST": "example.com"
}
}
}
}Replace ABS/PATH/TO with the real folder. fetch_url will refuse to run until MCP_FETCH_ALLOWLIST has at least one host.
Try a query
Once the client is connected, ask it to run:
SELECT id, title, body FROM notes WHERE id = ?with params [1]. You should get the welcome row from the sample database.
What this sample does not include
The full pack adds a TypeScript and Python starter with the guard layer (rate limit, optional audit log, structured errors), write-optional SQLite with the same SQL checks, a strict command allowlist (ls, read-only git, npm test), and a more complete fetch tool (redirect checks, HTML-to-Markdown). Those files are not in this repository.
If you want that command allowlist, the complete TypeScript / Python templates, or the optional AI IDE Rules Kit, look at the shop and X profile linked from the author's GitHub or X bio. The shop is dingster13.gumroad.com. This MCP kit may not be listed there yet — the sample is published first on purpose, so you can run it before anything is for sale.
License
MIT. See LICENSE.
Available Tools
2 toolsfetch_urlA
Fetch an allowlisted https URL and return up to 200 KB of text. Private networks are blocked.
Inputs: url (required).
sqlite_queryB
Run a single parameterized SELECT against the sample SQLite database. Writes are not available in this sample.
Inputs: sql (required), params.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 2 tools
The two tools have clearly distinct purposes: running a SQLite SELECT versus fetching an allowlisted URL. There is no realistic overlap or misselection risk between them.
Both names use snake_case, but the patterns differ: sqlite_query is resource_action while fetch_url is verb_noun. The names remain readable, but there is no consistent verb-first convention.
Two tools is thin even for a starter sample server. The scope is understandable, but the surface feels more like a minimal demo than a well-rounded tool set.
The SQLite tool is read-only and the URL tool only fetches, so write and broader lifecycle operations are absent. For a starter sample this may be intentional, but agents needing mutation or schema inspection would hit gaps.
Maintenance
Related MCP Connectors
Free bounded public HTTPS document and JSON reads. Exact host allowlist; real source provenance.
Security layer for AI agents: Retrieve for controlled SQL reads, Guard for allow/deny writes.
Safe, read-only Postgres and MySQL access for AI agents. Audit log + column-level controls.
Query 40 databases from Claude, ChatGPT, or Cursor — on any device. Read-only, encrypted, audited.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables safe, read-only SQL access to SQLite databases for AI agents, allowing schema exploration and SELECT queries with defense-in-depth protections.3MIT
- FlicenseNot gradedqualityBmaintenanceEnables read-only discovery and retrieval of allowlisted catalog datasets with host-assigned roles, strict tool validation, SQLite read-only enforcement, and audit hash chaining. It supports querying bounded daily aggregates while preventing arbitrary SQL, file access, shell commands, credential use, or role escalation.-
- AlicenseAqualityBmaintenanceEnables secure, read-only PostgreSQL queries, sandboxed file reads within a confined directory tree, and host CPU, memory, and disk metrics over stdio. It exposes these capabilities through narrowly scoped tools with strict validation and fail-closed guardrails.1MIT
- AlicenseNot gradedqualityCmaintenanceEnables exploring and querying a SQLite database loaded into memory through validated read-only SELECT/WITH statements, with enforced limits on database size, rows, result bytes, cells, and parameters.MIT