"Understanding How to Use a Browser" matching MCP servers:

A zero-dependency Node.js playground for debugging how MCP clients authenticate to an MCP server over HTTP.

Enables any agent to drive a dedicated Chromium instance through MCP, including opening and reading pages, taking steps, logging in via a password vault, downloading/uploading files, and replaying saved routines. It also gates sensitive actions behind approvals and can use client certificates or one-time card data for official procedures.

MCP server enabling Claude Code to capture screenshots of Windows applications through WSL2, allowing AI to diagnose UI errors and verify layouts.

Enables AI agents to call tools through a security gateway that intercepts requests, enforces deterministic policy and risk checks, blocks or escalates high-risk actions for human approval, and keeps tool credentials and audit logs protected.

Lets AI agents control a real Chrome browser through MCP, with scoped grants, per-site policies, a visible control overlay, an emergency stop, and human confirmation for privilege-raising actions.

Safety middleware for AI computer use agents that monitors, enforces policies, and provides guardrails to enable safe desktop interaction.

Enables a constrained subagent to perform foreground GUI actions on Hyprland—observing, focusing windows, clicking, typing, scrolling, dragging, and managing the clipboard—under per-action approval with lease, audit log, and kill-switch gating.

Enables an agent to drive a real browser while enforcing provenance-based gating so URLs and form values must trace to the user or an allowlist, never to untrusted page content.

Enables AI agents to request an independent second opinion from a different frontier model, with PII scrubbing before transit, live web search grounding, caching, and multi-provider failover.

This MCP server provides secure access to databases for AI agents, enforcing authentication, authorization, human approval, logging, and notifications to prevent dangerous actions.

Enables AI assistants to scan project dependencies and Infrastructure as Code files for security vulnerabilities and misconfigurations. It also provides automated fixing capabilities to remediate identified security issues.

Enables secure browser automation by letting agents fill Turnkey-stored secrets into web forms without the secret values ever appearing in the conversation, transcript, or model context.

Enables local AI agents on macOS to drive the user's own signed-in Chrome through their own isolated tabs, with per-client tokens, blocked site categories, masked secrets and human hand-offs. Agents can open and navigate tabs, read pages, send real clicks and typing, inspect network requests, and hand tabs back to the user, all without touching other tabs or stealing focus.

Rate-limits MCP tool calls with Upstash Redis, allowing a configurable number of invocations per sliding window and sharing counters across server instances.

Enables AI agents to control Tor Browser with full anonymity preservation, including navigation, Tor circuit management, and traffic interception.

Read-only observation of a single live URL: parses static HTML to report security posture, forms, links, accessibility signals, and leaks, with described fixes. SSRF-gated and safe, never executes JavaScript.

Enables AI agents to open and read pages that require login or heavy JavaScript by driving a real Chromium with a persistent profile kept across restarts, running headful on a virtual Xvfb display. Injects credentials from protected files without ever exposing cookie values, and adds domain allowlists, navigation throttling, and anti-bot cooldowns so agents back off before tripping risk controls.

Enables AI agents to drive a self-hosted Chromium through MCP tools for opening pages, reading content, taking screenshots, clicking, filling forms, and triggering final submissions, while a human watches the same browser live and can take over for logins or captchas. Server-side guardrails enforce per-session domain allowlists and denylists, refuse password, payment card, IBAN and one-time-code fields, isolate each token in its own browser context, and route submits, payments and orders through a separate final-action tool that can require human approval.

Enables AI agents to automate macOS desktop GUI applications through a policy-governed MCP proxy over Cua Driver, with fail-closed per-app allowlisting, a hardcoded sensitive-app denylist, an emergency kill-switch, masked audit logging, and self-healing app recovery with durable task journaling.

MCP server acting as a credential vault and browser execution broker, letting AI agents perform browser tasks (login, scraping, etc.) without exposing raw credentials, with per-tab billing and a monitoring UI.

An MCP server that gives AI clients controlled access to a real headless browser with domain allowlisting, SSRF protection, per-session isolation, audit logging, and persistent cookies/state across restarts.

Demonstrates Agent Passport System pre-dispatch authorization by issuing one-use grants and verifying signed authorization envelopes before proxied tool calls reach the upstream server.

Enables an MCP server to enforce External Verifier Contract authorization before proxied tool calls, allowing or denying requests based on a supplied bundle.

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.

Drives your own Chrome or Brave profile with cookies and extensions intact, so an MCP client can navigate, snapshot, click, type, screenshot and manage tabs, while logins and one-time codes are filled through the 1Password extension or CLI without the credentials ever reaching the model.

Provides agent certification and trust verification tools for AI agents, enabling certification checks, trust score calculations, audit ticket issuance, and emergency kill switch activation through the A-SOC trust network.

Retired. This MCP wrapper is no longer maintained. Check is now a Claude Code hook: npx @golproductions/check@latest --install. 120 free messages a day. Details: golproductions.com/check

An MCP server demonstrating OAuth 2.1 authentication with Scalekit, enabling per-user identity in MCP tools through JWT verification.

Enables coding agents to drive a browser through an authenticated session while being unable to read the credential that created it, by running the browser process under a separate OS user and stripping unsafe code-execution tools. The agent interacts over MCP with a hardened, tool-filtered automation server seeded from a login helper's saved session.

Enables AI agents to manage VPN infrastructure through provisioning, lifecycle control, traffic routing, and health monitoring.

MCP server that bridges authenticated sessions from a Chrome extension to CloakBrowser profiles via local WebSocket, enabling agents to sync, verify, and debug persistent login state without exposing raw credentials.

Enables Large Language Models to safely browse and interact with local file systems through secure directory listing, file reading, and content search capabilities. Built with comprehensive security controls and high-performance handling of large directories and files.

Enables MCP clients to triage SOC alerts while enforcing a trust firewall across retrieval, memory, and privileged actions. It exposes tools for alerts, logs, memory, and alert actions, with defenses that refuse risky operations in the presence of attacker-controllable content.

Enables executing arbitrary Joern/CPGQL queries against a loaded Code Property Graph, loading source code in multiple languages (PHP, Java, JavaScript, Python, C/C++) to build new CPGs, and inspecting loaded projects and the active workspace. It communicates with a running Joern server over its synchronous HTTP query endpoint, so no state polling is required.

Enables Claude to list Iceberg tables and run read-only DuckDB SQL queries, including time travel by snapshot, with per-caller token authentication and audit logging.

Enables AI agents to perform human-authorized, fully auditable GUI actions — clicking, typing, and window operations — on a local Windows desktop, gated by time- and count-limited leases that are denied by default until a person signs off in a popup. Every action is bound to a fresh screen capture, leaves screenshot-based evidence and decision logs on the local machine, and can be halted instantly via a panel button, hotkey, or kill-switch file.

Read-only ProofRelay MCP verifier for non-confidential evidence bundles. Exposes 22 public-safe tools, 11 resources, and 11 prompts for bundle integrity checks, receipt-chain review, checkpoint recommendations, MCP risk metadata review, and real-estate closing proof-pack readiness.

An MCP front-end for AnB. The tool surface is "use-don't-reveal": no tool returns a plaintext secret. Reveal paths require a TTY the server doesn't have, so the no-reveal guarantee is structural, not prompt-based — it holds even against a prompt-injected agent. Runs as a dedicated, narrowly-scoped AnB identity; exec is default-deny (only allowlist rules tagged scope=mcp run); command output

Enables AI agents to securely use real secrets (API keys, database passwords) by requiring human approval for each release, ensuring secrets never enter the model's context.

Scans suspicious messages, URLs, and text for scams inside any MCP-compatible AI assistant. No signup or API key needed for anonymous use.