MCP servers · least privilege

MCP servers with no destructive tools

3,151 MCP servers declare at least one tool and none of them is classed destructive: nothing deletes, overwrites or runs a command through a declared tool. 1,622 of them only read; the rest may create or edit.

Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
More
Cite

Cite this page

ModelsAtlas. “MCP servers with no destructive tools”. Retrieved 30 Oct 2026. https://modelsatlas.com/mcp/no-destructive/page/76

Verdicts · 3,151 servers

✓SAFE1,634!CAUTION522✕BLOCK995

Transports

A server can list more than one transport.

3,151 servers · 1,801–1,824 · ranked by trust score, then stars
U-he Preset Randomizer!CAUTION B

fannon/u-he-preset-randomizer

Generate u-he synth presets through randomization and merging of your existing presets.

★ 31GitHub stars2 writestdio
Robinhood!CAUTION B

rohitsingh-iitd/robinhood-mcp-server

The Robinhood MCP Server provides a comprehensive interface to the Robinhood Crypto API. This server handles authentication, account management, market data retrieval, and trading operations through both REST API and WebSocket interfaces.

★ 30GitHub stars4 write
fastify-mcp-server!CAUTION B

flaviodelgrosso/fastify-mcp-server

Wire MCP protocol in your Fastify applications with ease

★ 29GitHub stars7 readstreamable-http
Langchain MCP Tools!CAUTION B

hideya/langchain-mcp-tools-py

MCP to LangChain Tools Conversion Utility / Python

★ 29GitHub stars2 writestreamable-http
Unified Tool Graph!CAUTION B

pratikjadhav2726/unified-mcp-tool-graph

Instead of dumping 1000+ tools into a model’s prompt and expecting it to choose wisely, the Unified MCP Tool Graph equips your LLM with structure, clarity, and relevance. It fixes tool confusion, prevents infinite loops, and enables modular, intelligent agent workflows.

★ 27GitHub stars3 readstdio
VRChat OSC✓SAFE B

krekun/vrchat-mcp-osc

★ 27GitHub stars5 writestdio
Human!CAUTION B

olalonde/mcp-human

Human Assistance for AI Assistants

★ 26GitHub stars2 readstdio
BrowserLoop✓SAFE B

mattiasw/browserloop

A Model Context Protocol (MCP) server for taking screenshots and reading console logs of web pages using Playwright.

★ 23GitHub stars2 readstdio
@amplitude/mcp-analytics✓SAFE B

amplitude/Amplitude-MCP-Analytics-Node

Amplitude MCP Analytics SDK - MCP tracking for Amplitude Analytics. :sparkles: :calendar: Book a session with us :arrow_down:

★ 11GitHub stars8 readstdio
agentwheel!CAUTION B

NestDevLab/agentwheel

Skills, rules, and instructions across every AI agent. One source, every agent.

★ 9GitHub stars5 readstreamable-http
mcp-stdio✓SAFE B

shigechika/mcp-stdio

Bidirectional stdio ↔ HTTP gateway for MCP servers — connect clients to remote servers or publish local servers

★ 8GitHub stars1 writesse
jev-browser-control!CAUTION B

nexibeo/jev-browser-control

Let Claude code, chatgpt codex or control your own Chrome. Chrome extension + MCP server: Jev, TypeSafe's decision model, picks each click in ~0.5 s for a fraction of a cent. MIT, bring your own OpenRouter key.

★ 5GitHub stars2 write
vitest-agent✓SAFE B

spencerbeggs/vitest-agent

A Vitest reporter that keeps results of test runs focused for LLMs

★ 2GitHub stars2 read
llamator-mcp-server!CAUTION B

kyle122497/llamator-mcp-server

🚀 Automate LLM red teaming workflows with the MCP server for LLAMATOR, featuring asynchronous job handling and seamless integration.

★ 2GitHub stars2 writestreamable-http
Achiote!CAUTION B

KyaniteLabs/Achiote

Achiote — nostalgic food-memory reconstruction MCP server: recreates childhood dishes using ingredients available where you are

★ 1GitHub stars3 readstdio
Agentic Terminal Directory!CAUTION B

observer-protocol/at-directory

Agent commerce directory for Lightning, BOLT12, L402, and USDT merchants. Merchants self-register and land self-attested; an Observer Protocol attestation tier records what has been checked on each.

★ 1GitHub stars1 writestdio
ai_agents_api_library✓SAFE B

eater2/ai_agents_api_library

Verified catalog of 300+ APIs and MCP servers that AI agents can call after a one-time API key/OAuth setup. JSON, llms.txt and an MCP server for agents.

★ 1GitHub stars4 readstdio
browden!CAUTION B

nishantsny/browden

Read-only browser MCP server — a local, configurable MCP shell around real Chrome for LLM agents.

★ 1GitHub stars2 readsse
EDB_FILE_EXPLORER✓SAFE B

keyuraghao/EDB_FILE_EXPLORER

Cross-platform GUI, CLI and MCP server for Microsoft ESE databases (ntds.dit, SRUDB.dat, Exchange .edb, WebCacheV01.dat, Windows.edb) - digital forensics tool with multi-format extract and reporting

★ 1GitHub stars4 writestreamable-http
MCP_Detailed_With_Projects!CAUTION B

Birdieflemishspeaking787/MCP_Detailed_With_Projects

Learn MCP with practical projects for beginner to advanced use, connecting AI models, tools, and external systems

★ 1GitHub stars1 writestreamable-http
@orkestrel/probe✓SAFE B

orkestrel/probe

★ 0GitHub stars1 write
mcp✓SAFE B

kujolang/mcp

Kujo-native MCP server framework.

★ 0GitHub stars1 writestreamable-http
mcp-broker✓SAFE B

korals-ai/mcp-broker

An MCP gateway for tool servers that aren't up yet — upstreams appear mid-session instead of blocking agent startup

★ 0GitHub stars1 readstreamable-http
StoreTalk-MCP-Tiendanube!CAUTION B

Javi-cba/StoreTalk-MCP-Tiendanube

PROYECTO EN DESARROLLO: Gestioná tu tienda online conversando con tu asistente de IA. Creá, editá o borrá productos, actualizá precios y stock, revisá pedidos, consultá ventas y creá cupones en lenguaje natural, sin entrar al panel. Conectás tu tienda una vez y listo.

★ 0GitHub stars1 read

Other risk hubs

By job

Questions

What counts as destructive?
A tool that deletes, overwrites or runs commands. We class every tool the source declares as read, write or destructive; a server is listed here when it has tools and none is destructive.
How is this different from read-only?
Read-only servers have only read tools (1,622 servers). This list also includes servers whose tools write: 1,529 of the 3,151 do.
Does no destructive tools mean safe?
No. 1,634 are SAFE, 522 CAUTION and 995 BLOCK on this list. The verdict comes from the audit of the source, not from the tool list alone.

Keep going in your AI

Opens with this page as the source. We add nothing else to the prompt.