10 个#guardrails MCP 服务器和客户端

Hejdar

ARKALDA

Hejdar MCP Server — runtime policy enforcement for AI agents

Give your AI agents a wallet they can't abuse. Almega is an MCP server that puts a control layer in front of every payment: per-agent spending limits, allow-listed categories, 1-click human approval on sensitive transactions, and a full audit ledger. Two backends ship in one file

Waqi - AI Privacy Layer

ajprolific

Waqi — hosted MCP server that redacts PII from your business tools before the AI sees them. Audit log included.

Oi

carhaix

One shared AI setup for your team. Stop losing prompts in chats and docs. Save one shared setup, tailor it for your team, and use it across ChatGPT, Codex, Claude, Figma, and more.

Thinkneo Control Plane

thinkneo-ai

Enterprise AI Control Plane — governance layer between your AI applications and providers. Spend tracking, runtime guardrails, policy enforcement, budget limits, compliance readiness (SOC2/GDPR/HIPAA), and real-time provider health. Works with Claude, ChatGPT, Copilot, and any MC

FetchGate

MoAz06

The agent only answers from a page it actually read.

Impri

sekera-radim

Human-in-the-loop approval inbox for AI agents. An agent proposes an action, a human approves or rejects it, then it runs. Watchers + approval inbox, MCP + REST, self-hostable (MIT).

Runtime security oracle for AI agents — scan prompts, tool calls, and outputs before execution. Pay-per-call via x402/USDC, no accounts needed.

Local guardrail proxy for AI coding agents. Wraps any MCP server (stdio or Streamable HTTP) and blocks destructive tool calls — DROP TABLE, rm -rf, force-push — before they execute. TOFU catalog pinning against rug pulls, plus tool-poisoning and result-injection scanning. Single

A single `forecast` tool your agent calls before any irreversible action (send money, run SQL, delete data, post content). Returns a risk score (0–100), a reversibility class, named red flags from 28 failure modes, and a gate: proceed / confirm / human-required.