"Using Claude in Cursor to Run and Read Outputs from PowerShell or Terminal" matching MCP servers:

Your AI agent can't touch Stripe, Gmail, Slack or more, without your tap. mild.run intercepts every sensitive action and sends you a notification to approve or block — before anything executes. Agent never holds credentials.

A secure terminal execution server that enables controlled command execution with security features and resource limits via the Model Context Protocol (MCP).

A secure MCP server example that demonstrates how to deploy to Google Cloud Run with authentication and identity token protection. Serves as a tutorial template for building production-ready MCP servers in the cloud.

Enables Claude AI to play a cookie-themed dungeon RPG, including clicking cookies, recruiting heroes, exploring dungeons, and managing inventory through natural language commands.

A secure email gateway MCP server that protects AI agents from prompt injection attacks by scanning emails for malicious instructions before they reach the agent.

Encrypted token store for Claude Code sessions, providing MCP tools for secure secret management with macOS Keychain integration, per-project allowlists, and native dialog input.

A secure MCP server that allows Claude to read and write local files on your machine with explicit approval gating for each access.

Deterministic pre-execution safety certification for autonomous AI agents. Evaluates proposed actions across scope, reversibility, and sensitivity constraints — returns Certified, Warning, Escalate, or Blocked decisions with cryptographic audit trails. 3 tools: certify_action, check_budget, get_certification_history. Built on the QAE Safety Kernel (Rust + PyO3 bindings, pip install qae-safety).

An MCP server that sends plain-text Gmail messages only to a predefined allowlist, using generated recipient IDs instead of email addresses to enforce security.

A proof-of-concept attack that exploits Model Context Protocol (MCP) tool registration to achieve persistent agent poisoning in AI assistants like Cursor, embedding malicious instructions that persist across chat contexts without requiring tool execution.

Enables authorized web/API security analysis as a stateful, multi-stage workflow with persistent session state, human-controlled validation, candidate versus confirmed finding tracking, and stop conditions.

Provides an OAuth 2.1 authorization and login layer plus tool registration for MCP connectors, so a connector can be added to Claude clients with only a URL and password. It also shapes tool names, descriptions and input schemas for what Claude clients actually display, returning compact results and readable error messages.

Enables a sandboxed Claude Code session to see editor selections, review proposed edits as VS Code diffs, and read workspace diagnostics through an MCP-over-WebSocket IDE bridge.

Enables analytics agents to run validated read-only SQL against a warehouse API with enforced row limits, required JSON responses, and secret-safe audit metadata.

Enables named, authenticated callers to run read-only introspection of a Google Cloud project and to read and write per-tenant notes that stay isolated between callers. Every request is scope-checked against a reviewed policy, rate-limited per caller, and audited without ever logging argument values.

Enables auditing public GitHub repositories over MCP Streamable HTTP, including classifying URLs, cloning repositories, scanning manifests, CI workflows, and credential signals, and producing findings, file trees, summaries, and PDF reports.

Enables autonomous agents and MCP-compatible clients to enforce policy-approved actions using risk tiers, reversible checkpoints, and token approval bounds.

Enforces Awaken Church's governance policies by scanning content, correcting language, blocking unauthorized integrations, routing projects, and flagging content for review.

Enables local spend caps and Ed25519-signed provenance receipts for AI agent tool calls, allowing any MCP host to gate model usage and independently verify signed decisions without sending prompts or content.

Blocks AI coding assistants from installing non-existent packages and accessing non-existent files.