MCP Security Guides

Security basics

ChatGPT Skills, Perplexity Skills and Claude Skills: How to Install Each and What Differs (Nov 2026)

ChatGPT, Perplexity and Claude all support skills: a folder with a SKILL.md file that teaches the assistant one task, in the same open Agent Skills format. Which plans have them, how to install, upload and remove a skill in each app, how they differ, and five checks before you install one.

12 min read Read Security basics

Best MCP Servers for Claude Code: Graded Picks by Job (Nov 2026)

Add one MCP server per job you do every week: GitHub or GitLab for code, Context7 for library docs, your database's own server, Sentry for errors. 25 graded picks, the claude mcp add line for each where we have one, and two servers to skip.

10 min read Read Security basics

MCP Logging and Audit Logs: Record What Your AI Agent Did (Nov 2026)

MCP keeps no audit log for you, and its own logging feature is deprecated. What a log of agent tool calls should record, why each agent needs its own identity, Claude Code's OpenTelemetry events and hooks, vendor audit logs, and what's left when none of them exist.

13 min read Read Security basics

MCP Streamable HTTP: How It Works, vs stdio and SSE (Nov 2026)

MCP Streamable HTTP is the standard transport for MCP servers you reach by URL: one endpoint, one POST per message, JSON or a short SSE stream back. How it works under the 2026-07-28 spec, what changed (no sessions, no GET stream), how it differs from stdio and the old HTTP+SSE transport, client setup, error codes, security, and a 5-step move off SSE.

16 min read Read Security basics

OpenAPI to MCP: Turn a Swagger Spec Into a Lean MCP Server (Nov 2026)

How to turn an OpenAPI or Swagger spec into an MCP server: the generators (FastMCP, openapi-mcp-generator, AWS Labs, Speakeasy, FastAPI-MCP and gateways), why one tool per endpoint floods the model's context, a 6-step trim list, Swagger 2.0 tips, security checks, and mcpo for the reverse trip, MCP to OpenAPI.

12 min read Read Security basics

Python MCP SDK vs FastMCP: Which to Use and How to Log (Nov 2026)

The official Python MCP SDK (`mcp` 2.3) and FastMCP 4 (`fastmcp`) side by side: what FastMCP is, which one to pick, the v2 rename behind "No module named 'mcp.server.fastmcp'", transports, logging to stderr instead of stdout, and the versions that fix 2026's security advisories.

12 min read Read Security basics

Claude Code Settings: settings.json, Scopes and Permissions (Nov 2026)

Claude Code settings live in JSON files at four levels: user (~/.claude/settings.json), shared project (.claude/settings.json), project local (.claude/settings.local.json) and managed. Which file wins, how allow, ask and deny permission rules work, and where MCP, skills and subagent config lives.

12 min read Read Security basics

Claude Code Skills: What They Are, How to Install Them and How to Check One (Nov 2026)

A Claude Code skill is a folder with a SKILL.md file: instructions, plus optional scripts, that Claude loads when your request matches it or when you type /skill-name. How skills load, how to install and manage them, where to find them, and seven checks before you install one.

15 min read Read Security basics

Claude Code Subagents: How to Create, Scope and Use Them (Nov 2026)

A Claude Code subagent is a helper with its own context window, prompt, tools and model, defined in a Markdown file in .claude/agents/. Where the files live, every front matter field, how to limit tools and MCP servers, and when to use a subagent, a skill or an MCP server.

12 min read Read Security basics

Local MCP Server: stdio vs Localhost, Open Ports and Safe Setup (Nov 2026)

A local MCP server runs on your own machine. Most use stdio and open no port at all; some open a local HTTP port. How to tell which you have, how to see what's listening, why 127.0.0.1 beats 0.0.0.0, what DNS rebinding is, and how to add a token or use a tunnel instead of an open port.

12 min read Read Security basics

MCP Interview Questions and Answers: 32 Questions Checked Against the Spec (Nov 2026)

32 Model Context Protocol interview questions with short answers checked against the current 2026-07-28 spec: basics, messages, transports, tools vs resources, OAuth, security and server design. Each answer links to a longer explainer, and the old 2025 answers are flagged where they changed.

13 min read Read Security basics

MCP TypeScript SDK Vulnerabilities: Are You Affected? (Nov 2026)

Seven advisories hit the official MCP TypeScript SDK between 30 September and 15 October 2026; NVD published four of the CVEs on 5 November. Who's affected, which versions fix them, the four fixes that need a code change, how to find old SDK copies in your npx cache, and where the servers we list stand.

10 min read Read Security basics

MCP Sandbox: How to Sandbox MCP Servers on Your Machine (Nov 2026)

A local MCP server runs with your full rights, and most AI clients don't sandbox it. Three levels of isolation (a separate user, a container with no host disk, a network allowlist), which servers need the network at all, a small proxy that refuses write calls, and what a sandbox still can't stop.

14 min read Read Security basics

Skill Creator: Build, Test and Check Your Own Claude Skill (Nov 2026)

skill-creator is Anthropic's open-source skill for building other skills. Install it in Claude Code, describe what you want, and it drafts the SKILL.md, tests it against a no-skill baseline and tunes the description. Step by step, plus a security checklist for what you ship.

12 min read Read Security basics

What Is an MCP Server? MCP Meaning, How It Works and Examples (Nov 2026)

An MCP server is a small program that gives an AI app like Claude, Cursor or VS Code new tools and data. It lists what it can do, the AI model picks a tool, and the app runs it with your approval. MCP (Model Context Protocol) is the open standard they all speak.

10 min read Read Security basics

How to Limit MCP Tools: Allowlists, Permissions and Read-Only Mode (Nov 2026)

Hiding an MCP tool and blocking it are different things. Which settings in Claude Code, Cursor and VS Code only hide a tool, which ones refuse the call, and why a read-only flag and a narrow token are the limits that hold.

11 min read Read Security basics

MCP Server Not Working? A 5-Minute Checklist and Error Fixes (Nov 2026)

Most broken MCP servers fail in one of six places: the command won't start, the handshake fails, tools/list comes back empty, sign-in, a timeout, or a tool error. Check them in that order. Includes what error -32000 means and fixes for Claude Desktop, Claude Code and Cursor.

14 min read Read Security basics

Claude Code MCP Server Setup: claude mcp add, list, Config and Fixes (Nov 2026)

Add a server to Claude Code with claude mcp add (HTTP or stdio), pick a scope (local, project .mcp.json, or user), check it with the claude mcp list command and /mcp, then limit its tools with mcp__server__tool permission rules.

14 min read Read Security basics

MCP Gateway: What It Is, When You Need One and How to Choose (Nov 2026)

An MCP gateway is a proxy that sits between your AI apps and many MCP servers. It gives everyone one endpoint and puts sign-in, tool allow-lists, logs, rate limits and secrets in one place. What it adds, when you need one, how to choose one (open-source and managed options), and what it can't stop.

10 min read Read Security basics

MCP Inspector: How to Run It and Use It to Check a Server (Nov 2026)

The MCP Inspector is the official tool for testing and debugging MCP servers. Run it with npx as a web UI or CLI, connect a local or remote server, and read its tools, annotations and sign-in before an agent uses it.

12 min read Read Security basics

MCP vs API: What MCP Adds and When a Plain API Is Better (Nov 2026)

MCP doesn't replace your API. It sits on top of it: a server lists tools with names, descriptions and input schemas, and any MCP client can hand them to a model that decides which to call. Use a plain API when your code already knows the call.

11 min read Read Security basics

How to Create an MCP Server, Securely: Python and TypeScript Step by Step (Oct 2026)

Create an MCP server with an official SDK: define a few narrow tools with input schemas, run it over stdio, test it with the MCP Inspector, then work through the security checklist before anyone else connects to it.

6 min read Read Security basics

MCP Prompt Injection: How the Attacks Work and How to Defend Against Them (Oct 2026)

Prompt injection in MCP is text from a tool result, a resource or a tool description that the model follows as if it were your instruction. No server update fixes it. You limit the damage: approval on actions, narrow tokens, and never mixing untrusted content, private data and outbound tools in one session.

6 min read Read Security basics

MCP Resources vs Tools: What's the Difference and When to Use Each (Oct 2026)

In MCP, tools are actions the model decides to call; resources are data the app or user attaches as context. Tools can change things, so they need approval. Resources are read-only, but their content is still untrusted input.

5 min read Read Security basics

How to Keep API Keys Out of Your MCP Config (Claude Code, Cursor, VS Code) (Oct 2026)

Don't paste API keys into mcp.json or .mcp.json. Reference an environment variable or a prompted input instead, keep the real value outside the project, and prefer servers that sign in with OAuth and scoped tokens.

5 min read Read Security basics

MCP Authentication and MCP OAuth: How Sign-In Works (Nov 2026)

Remote MCP servers sign you in with OAuth 2.1; local servers read credentials from the environment. How MCP OAuth works under the 2026-07-28 spec, what changed, how to add authentication to your own MCP server, why your client keeps asking you to sign in, and which grants are too broad.

12 min read Read Security basics

MCP Security Roundup (Oct 2026)

October's verified MCP security news: advisories for VS Code and n8n, a run of fixes in the official TypeScript, Python and Java SDKs, Stripe's 30 November key deadline, and the end of Salesforce's local MCP server on 2 November.

7 min read Read Security basics

MCP Tool Annotations Explained: readOnlyHint, destructiveHint and Why They're Only Hints (Nov 2026)

MCP tool annotations tell a client whether a tool reads, writes, deletes or reaches the outside world. They are hints from the server, not guarantees, so they can't replace your own review, especially after a server update.

5 min read Read Security basics

MCP Tool Call Timeout: Why Long Tools Fail and How to Fix It (Nov 2026)

Long MCP tool calls usually die on an idle timer, not a hard limit. In Claude Code, raise the per-server timeout or the idle window; on the server, send progress notifications or return a task instead of blocking.

6 min read Read Security basics

MCP Tool Errors: Why Your Agent Keeps Calling a Failing Tool (Nov 2026)

An agent retries a failing MCP tool because the error reaches the model as text, and nothing in that text says the failure is permanent. Servers should return isError with plain advice on whether a retry can help.

7 min read Read Security basics

Best MCP Server for SEO: Search vs. SEO-Data Servers, Compared (Oct 2026)

For live web search, use Tavily or Exa (both A-graded). For keyword volumes, difficulty and backlinks, you need an SEO-data server: Ahrefs grades A, Semrush and DataForSEO B, Ubersuggest C. Most SEO work needs one of each.

4 min read Read Security basics

Why Your MCP Servers Eat Your Context Window (and How to Get It Back) (Oct 2026)

Every connected MCP server puts its tool descriptions into the model's context before you type anything. Measure what each one costs, turn off the ones you don't need, and filter the rest down to the tools you actually use.

4 min read Read Security basics

The OWASP MCP Top 10: Risks and How They Map to a Trust Score (Sep 2026)

The ten security risks that matter most for Model Context Protocol servers, drawn from OWASP's LLM Top 10 and MCP security research - and how each maps to the four signals TrustedMCP scores.

2 min read Read

A Atlassian Rovo MCP Server

Atlassian MCP Server: Setup, Rovo Permissions and Security (Oct 2026)

How to connect Atlassian's official Rovo MCP server to Jira, Confluence, Compass and more, how OAuth 2.1 and API tokens work, which tools admins must turn on, why the 2026 MCP Atlassian CVEs belong to a different project, and why it grades A.

6 min read Read C browser-use MCP

browser-use MCP Server: Setup, Safety and Playwright Comparison (Nov 2026)

browser-use's MCP server gives an AI client a real browser and an optional autonomous agent. How to set it up, whether you need an OpenAI key, how to restrict domains, which mode runs code, why it grades C, and how it compares with Playwright and Chrome DevTools.

11 min read Read C BrowserTools MCP

BrowserTools MCP: Setup, the 1.x Flaw and Safety Grade (Nov 2026)

What AgentDesk's BrowserTools MCP reads from your Chrome, how it differs from browser-use, how to set up 2.x in Cursor, Claude Code and Claude Desktop, why every 1.x install must upgrade, its two advisories, and why it grades C.

11 min read Read C Burp Suite MCP Server

Burp MCP Server: PortSwigger Setup, Approvals and Risks (Nov 2026)

PortSwigger's MCP Server extension connects Burp Suite to Claude, Cursor or VS Code. How to install it, which actions ask you first and which don't, what can leak from proxy history, and why it grades C.

9 min read Read C Chrome DevTools MCP

Chrome DevTools MCP Server: Setup, Safe Flags and Why It Grades C (Oct 2026)

How to set up Google's Chrome DevTools MCP server in Claude Code, Cursor and VS Code, the flags that make it safer (isolated profile, no JavaScript evaluation, URL allow-lists), its four fixed advisories, and how it compares with Playwright MCP.

5 min read Read B Clay MCP Server

Clay MCP Server: Setup in Claude, Credits and Safety (Nov 2026)

Clay MCP is Clay's official hosted server for finding and enriching people and companies from Claude, ChatGPT and other AI tools. How to connect it, what each search costs in credits, which admin controls matter, how it differs from the old Clay personal CRM (now Mesh), and why it grades B.

10 min read Read A Cloudflare API MCP Server

Cloudflare MCP Server: Setup, Code Mode and Security (Oct 2026)

How to connect Cloudflare's official API MCP server, how code mode works, which older Cloudflare servers it replaces, and why it grades A.

6 min read Read A Context7 MCP

Context7 MCP Server: Setup, API Key and Security (Oct 2026)

How to connect Upstash's Context7 MCP server to Claude Code, Cursor and VS Code, when you need an API key, where to keep it, what the ContextCrush CVE taught, and why it grades A.

5 min read Read C DBHub

DBHub MCP Server: Setup, Read-Only Mode and the MySQL Advisory (Nov 2026)

DBHub is Bytebase's open-source MCP server for Postgres, MySQL, MariaDB, SQL Server, Oracle and SQLite. How to set it up, how to make it truly read-only, who the MySQL/MariaDB read-only advisory affects and what to do, how to run it over HTTP safely, and why it grades C.

11 min read Read B Docker MCP Gateway (MCP Toolkit)

Docker MCP Gateway and Toolkit: Setup, Catalog and Security (Oct 2026)

How Docker's MCP Gateway, MCP Toolkit and MCP Catalog fit together, how to connect Claude Code, Claude Desktop, Cursor and VS Code, what the containers isolate and what they don't, and why the Gateway grades B while every server it runs still needs its own review.

6 min read Read D DuckDuckGo MCP Server

Web Search MCP Servers Compared: Free and Paid, Graded (Nov 2026)

Six web search MCP servers side by side: hosted Tavily, Exa and Perplexity, Brave's local server, and the free DuckDuckGo and SearXNG options. What each needs, how to set it up, and the risks behind each grade.

8 min read Read A Exa MCP

Exa MCP Server: Setup, OAuth vs API Key, Tools and Safety (Nov 2026)

Exa's official MCP server in plain terms: the hosted URL, how to connect it to Claude Code, Cursor or VS Code, keyless vs OAuth vs API key, which tools to turn on, what agent_run can spend, and why it grades A.

9 min read Read B Figma MCP Server

Figma MCP Server: Setup, Write Tools and Security (Oct 2026)

How to connect Figma's official MCP server to Claude Code, Cursor and VS Code, what its write tools can do, and why it grades B.

5 min read Read D Filesystem MCP Server

Filesystem MCP Server: Setup, Read-Only Mode and Security Risks (Nov 2026)

The official reference server gives an AI client file access inside folders you choose. How to set it up, when to use --read-only and --confine-roots, what its six path-escape advisories mean, the open Windows gap, and why it grades D.

11 min read Read B Firecrawl MCP Server

Firecrawl MCP Server: Setup, API Keys and the Open CVE (Oct 2026)

How to connect Firecrawl's official MCP server for scraping, crawling, search and document parsing, how to handle your API key, what the open firecrawl_parse CVE means for local setups, and why it grades B.

6 min read Read A GitHub MCP Server

GitHub MCP Server: Setup, Tokens and Security (Oct 2026)

How to connect GitHub's official MCP server to Claude Code, Cursor, VS Code and Claude Desktop, which token to use, how to lock it down, and why it grades A.

5 min read Read D Godot MCP Server (Coding-Solo/godot-mcp)

Godot MCP Server: Setup in Claude Code and Cursor, and What It Can Change (Nov 2026)

Godot has no official MCP server. How to set up Coding-Solo's godot-mcp, the one most people use, in Claude Code, Claude Desktop and Cursor, why to build it from source instead of npx, what it can run and write in your project, and why it grades D.

9 min read Read A Hugging Face MCP Server

Hugging Face MCP Server: Setup, Spaces and Security (Oct 2026)

How to connect the official Hugging Face MCP server at huggingface.co/mcp, the difference between anonymous, OAuth and token access, what the OAuth sign-in asks for, why Spaces and Jobs need care, and how it differs from the Hugging Face MCP course.

5 min read Read A Linear MCP Server

Linear MCP Server: Setup and Security (Sep 2026)

How to connect the official Linear MCP server to Claude, Cursor, and VS Code, and why its OAuth-over-HTTP model makes it an A-grade listing.

2 min read Read D markdownify-mcp

markdownify-mcp: Don't Install It Until It's Patched (Oct 2026)

markdownify-mcp converts files and web pages to Markdown, but three published vulnerabilities still affect its latest version. It grades D. Here's what's wrong and what to do instead.

2 min read Read D MCP for Blender (ahujasid/mcp-for-blender, formerly blender-mcp)

Blender MCP Server: Setup, the Addon, and Why It Grades D (Nov 2026)

Blender MCP (ahujasid/mcp-for-blender) lets Claude, Cursor or Codex build 3D scenes by running Python in your live Blender. How to set up the server and the addon, why it grades D, and how to run it safely.

9 min read Read B Microsoft 365 MCP Server (Softeria ms-365-mcp-server)

Outlook MCP Server: Setup for Claude, Read-Only Mode and Safety (Nov 2026)

The most used Outlook MCP server is Softeria's open-source ms-365-mcp-server. Add it to Claude Code or Claude Desktop with one npx line, sign in with Microsoft, and use --preset outlook and --read-only to keep it to mail and calendar. Your options, setup and its grade (B).

6 min read Read B Miro MCP Server

Miro MCP Server: Setup, Team Access, Tools and Safety (Nov 2026)

Miro's official MCP server in plain terms: the hosted URL, how to connect Claude Code, Cursor, VS Code or Claude, how team selection limits what the agent sees, what changed with the canvas tools, and why it grades B.

11 min read Read A MongoDB MCP Server

MongoDB MCP Server: Setup, Read-Only Mode and Security (Oct 2026)

How to connect MongoDB's official MCP server to Claude Code, Cursor and VS Code, how to lock it to read-only with a least-privilege user, and why it grades A.

5 min read Read A n8n MCP Server (instance-level)

n8n MCP Server: Built-in Server vs n8n-nodes-mcp, Setup and Security (Oct 2026)

Three different things are called n8n MCP. Which one to use, how to turn on n8n's built-in MCP server safely, how to expose only chosen workflows, and why it grades A.

6 min read Read B Notion MCP

Notion MCP Server: Setup, Permissions and Security (Oct 2026)

How to connect Notion's official hosted MCP server to Claude Code, Cursor and VS Code, what it can read and change, and why it grades B.

5 min read Read A Perplexity MCP Server

Perplexity MCP Server: Setup, API Keys and Security (Oct 2026)

How to connect Perplexity's official MCP server, hosted with OAuth or an API key or locally with npx, what its four tools do, how billing works, and why it grades A despite org-wide keys with no scopes.

5 min read Read C Pexels MCP Server (mcp-pexels)

Pexels MCP Server: Setup, Free API Key and Safe Use (Nov 2026)

Pexels has no official MCP server. How to set up the community mcp-pexels server in Claude Code, Claude Desktop and Cursor, get the free Pexels API key and keep it in an env var, stay inside the rate limits and licence, and why it grades C.

8 min read Read C Playwright MCP

Playwright MCP Server: Setup, Safe Flags and Security (Oct 2026)

How to set up Microsoft's Playwright MCP server in Claude Code, Cursor, VS Code and Claude Desktop, how to run it safely, and why it grades C.

6 min read Read B Product Hunt MCP Server

Product Hunt MCP Server: Setup, Launch Tools and Safety (Nov 2026)

Product Hunt's official hosted MCP server in plain terms: the URL, how to connect Claude, ChatGPT, Claude Code or Cursor, the 17 tools, what signing in lets it change on your launches, how the preview-and-confirm step really works, and why it grades B.

13 min read Read B ProxmoxMCP-Plus

Proxmox MCP Server: Setup, Safe API Tokens and Our Grade (Nov 2026)

The most used Proxmox MCP server is ProxmoxMCP-Plus. Install it with uvx, give it its own Proxmox API token with the smallest role that works, and turn on approval for deletes and rollbacks. Setup for Claude Code, Claude Desktop and Cursor, plus why it grades B.

8 min read Read A Railway MCP (railway mcp in the Railway CLI)

Railway MCP Server: Setup, Claude Code and Moving Off @railway/mcp-server (Nov 2026)

Railway's MCP server now lives in the Railway CLI. Run railway mcp install to set it up for Claude Code, Cursor and other tools, or use OAuth at mcp.railway.com. How to move off the deprecated @railway/mcp-server package, what the tools can change, and why it grades A.

6 min read Read B Sanity MCP Server

Sanity MCP Server: Setup, Permissions and What It Can Publish (Nov 2026)

Sanity's official hosted MCP server, in plain terms: how to connect it to Claude Code, Cursor or VS Code, how to make a read-only token, which tools can change live content or open up a dataset, and why it grades B.

12 min read Read A Sentry MCP Server

Sentry MCP Server: Setup and Security (Oct 2026)

How to connect Sentry's official MCP server to Claude Code, Cursor, and VS Code, how to run it against self-hosted Sentry, and why it earns a full A on TrustedMCP.

5 min read Read C SerpApi MCP Server

Google Patents MCP Server: Which One to Use and How to Set It Up (Nov 2026)

There is no official Google Patents MCP server. The package called google-patents-mcp hasn't had a release since April 2025. Here is the maintained way to search Google Patents from Claude, Cursor or VS Code, what it costs, the other patent servers worth knowing, and how the one we graded scores.

11 min read Read A Slack MCP Server

Slack MCP Server: Setup, Scopes and Security (Oct 2026)

How to connect Slack's official hosted MCP server to Claude Code, Claude Desktop, Cursor and Codex, what its OAuth scopes cover, and why it grades A.

5 min read Read A Splunk MCP Server

Splunk MCP Server: Setup, Tokens and Which Tools to Allow (Nov 2026)

What Splunk's official MCP server app (Splunkbase 7931) does, how to install it, create encrypted tokens and connect Claude Code, Cursor or Claude Desktop, which roles and tools to give an agent, its three past advisories, and why it grades A.

8 min read Read B Stack Overflow MCP Server

Stack Overflow MCP Server: Setup and Trust Assessment (Oct 2026)

How to connect the official Stack Overflow MCP server to an agent, what it can and cannot do, and why it grades B.

2 min read Read A Stripe MCP

Stripe MCP Server: Setup and Security (Oct 2026)

How to connect Stripe's official hosted MCP server to Claude Code, Cursor, and VS Code, how to keep it safe with OAuth, agent API keys and sandboxes, and why it grades A.

5 min read Read A Supabase MCP Server

Supabase MCP Server: Setup, Scoping and Security (Oct 2026)

How to connect Supabase's official MCP server to Claude Code, Cursor and VS Code, how to lock it to one project in read-only mode, and why it grades A.

5 min read Read A Tavily MCP

Tavily MCP Server: Setup, OAuth vs API Key, and Security (Oct 2026)

How to connect Tavily's official MCP server for web search, extraction, crawling and research, why you should use OAuth instead of putting your API key in the URL, and why it grades A.

4 min read Read A VS Code MCP

VS Code MCP Server: Setup and Security (Sep 2026)

How to enable MCP servers in VS Code, and what the editor does and does not secure for you.

1 min read Read B X (Twitter) MCP Server

Twitter (X) MCP Server: Official Setup, API Costs and Posting Risks (Nov 2026)

X now runs its own MCP server at api.x.com/mcp. How to connect it to Claude, Cursor or VS Code with the xurl bridge, what each call costs on X's pay-per-use API, what can go wrong when an agent posts as you, and why it grades B.

14 min read Read C Yahoo Mail MCP Server (local-mcp-server-yahoo-mail)

Yahoo Mail MCP Server: Setup, App Password and Safe Use (Nov 2026)

Yahoo has no official MCP server. How to set up the community Yahoo Mail MCP server (local-mcp-server-yahoo-mail) in Claude Code and Claude Desktop with a Yahoo app password and Yahoo's IMAP settings, keep it read-only or drafts-only, block send with deny rules, handle prompt injection from incoming mail, and why it grades C.

10 min read Read B Zapier MCP

Zapier MCP Server: Setup, Task Costs and Security (Oct 2026)

How to connect Zapier MCP in Claude Code, Claude, Cursor and VS Code, the difference between agentic and managed mode, what each tool call costs in tasks, how to lock action fields, and why it grades B.

6 min read Read