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.
Cite
Cite this page
ModelsAtlas. “MCP servers with no destructive tools”. Retrieved 30 Oct 2026. https://modelsatlas.com/mcp/no-destructive/page/91
Verdicts · 3,151 servers
Transports
- stdio, runs locally2,195
- Streamable HTTP, remote1,064
- SSE, legacy remote253
- Not declared563
A server can list more than one transport.
hmmhmmhm/daiso-mcp
다이소를 MCP로! 뿐만 아니라 한국 로컬 리테일과 영화관 조회를 MCP, CLI, Codex Skill로 가능하게 해줍니다.
sveltejs/mcp
The official svelte MCP for all your agentic needs.
neon-solutions/add-mcp
The open MCP config tool - npx add-mcp
cyberchitta/llm-context.py
Share code with LLMs via Model Context Protocol or clipboard. Rule-based customization enables easy switching between different tasks (like code review and documentation). Includes smart code outlining.
vxasi/blender-mcp-vxai
chrrxs/robloxstudio-mcp
MCP server for Roblox Studio runtime debugging, playtest control, screenshots/input, multiplayer testing, and per-peer server/client eval from AI agents.
ivo-toby/mcp-openapi-server
MCP Server (Model Context Protocol) for turning OpenAPI specifications into a MCP Resource
mjyoo2/overleafmcp
Model Context Protocol (MCP) server that lets AI assistants read Overleaf projects, parse LaTeX document structure, and push section-level edits back via Git. Compatible with Claude Desktop, Cursor, Windsurf, and any MCP client.
zenolab124/monet
Monet — Multi-engine mission control for coding agents (Claude Code and Codex today). Browse, search, and drive your agent sessions from a native desktop app.
enuno/unifi-mcp-server
An MCP server that leverages official UniFi API
jae-jae/g-search-mcp
A powerful MCP server for Google search that enables parallel searching with multiple keywords simultaneously.
tranhoaihung/figma-ui-mcp
AI can draw UI directly on the Figma canvas via JavaScript, and read existing designs back as structured data. Works with Claude Code, Cursor, VS Code, and Windsurf. No API KEY required.
dawnslab/jebmcp
flankerhqd/jebmcp
api-200/api200
API 200 is an open source API gateway to simplify 3rd-party integrations. Import endpoints, set up caching, retries, and mocks. Access all services via one URL. Monitor logs, track errors, and get alerts on API incidents.
praneshasp/foundry-mcp-server
An experimental MCP Server for foundry built for Solidity devs
nhevers/moltbrain
Long-term memory layer for OpenClaw & MoltBook agents that learns and recalls your project context automatically.
kronova-intelligent-systems/asset-intel-orchestration-engine
Open-source platform for multi-agent AI orchestration, real-world asset tokenization, workflow automation. Includes OAuth 2.1 MCP agent management, a 44-field RWA schema, Stripe, designed to integrate with AetherNet QUAS & KVS sovereign, post quantum, architecture agnostic omni-substrate, semantic m
dmontgomery40/mcp-3d-printer-server
Connects MCP to major 3D printer APIs (Orca, FULU's Orca/Bambu, OctoPrint, Klipper, Duet, Repetier, Prusa, Creality). Control prints, monitor status, and perform advanced STL operations like scaling, rotation, sectional editing, and base extension. Includes slicing and visualization.
zhixiaoqiang/antd-components-mcp
An MCP service for Ant Design components query | 一个减少 Ant Design 组件代码生成幻觉的 MCP 服务,包含系统提示词、组件文档、API 文档、代码示例和更新日志查询
formulahendry/mcp-server-code-runner
Code Runner MCP Server
montevive/penpot-mcp
Penpot MCP server
cyanheads/git-mcp-server
Clone, commit, diff, branch, merge, rebase, cherry-pick, stash, tag, and manage worktrees in local git repos via MCP. STDIO or Streamable HTTP.
hverlin/mise-vscode
VS Code extension for mise-en-place (https://mise.jdx.dev/)
Other risk hubs
By job
Questions
What counts as destructive?
How is this different from read-only?
Does no destructive tools mean safe?
Keep going in your AI
Opens with this page as the source. We add nothing else to the prompt.
- What counts as destructive?
- How is this different from read-only?
- Does no destructive tools mean safe?