# Zapier MCP: Connecting AI Clients to Thousands of Apps via the Model Context Protocol > Zapier MCP is a hosted Model Context Protocol server that gives any MCP-compatible AI client governed access to over 9,000 applications, with SOC 2 Type II certification and no code or infrastructure required from the user. This repository distributes the plugin that installs into AI clients to add onboarding, a first-use demo, and role-specific skills on top of the base connection. **zapier/zapier-mcp** — Official plugin distribution for the hosted Zapier MCP server. Install it in your AI client and connect to thousands of apps. - Repository: https://github.com/zapier/zapier-mcp - Website: https://docs.zapier.com/mcp - Stars: 420 · Forks: 68 - Language: Unknown - License: MIT - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/zapier-zapier-mcp ## What MCP Is and What Zapier MCP Adds to It The Model Context Protocol is an open standard for connecting AI language models to external tools and data sources. An MCP server exposes a set of callable tools over a standard transport layer, and any client that speaks MCP can invoke those tools without custom integration code. Zapier MCP is a hosted MCP server run by Zapier. When connected, it gives an AI client access to the automation actions and triggers Zapier has built for over 9,000 applications. From an AI client connected to Zapier MCP, a user can ask the assistant to send an email via Gmail, create a card in Trello, pull records from a spreadsheet, or trigger any Zapier workflow, all in plain English. The README describes it as governed access: Zapier applies permissions and routing rather than giving the AI client direct API access to each connected service. The zapier-mcp repository itself is the plugin distribution, not the server. The server runs at mcp.zapier.com. The repository holds the plugin that lives in supported AI clients and layers guided onboarding, a first-use demo, and role-tailored skills on top of the raw server connection. ## How Zapier MCP Connects to Supported AI Clients Connecting Zapier MCP to any MCP-compatible client requires only a server URL. The README directs users to docs.zapier.com/mcp/clients for step-by-step client-specific instructions. For VS Code, the connection is a direct HTTP MCP configuration pointing to https://mcp.zapier.com/api/v1/connect with type: http. Four clients also install the plugin from this repository, which adds onboarding and role-tailored skills on top of the base connection: Claude Code pulls the plugin from the zapier/marketplace or anthropics/claude-plugins-official repositories. Cursor installs it via the badge link in the README or from cursor/mcp-servers. OpenAI Codex and GitHub Copilot CLI both use zapier/marketplace. Claude Cowork uses anthropics/knowledge-work-plugins. For Gemini CLI, the installation command is: ```bash gemini extensions install https://github.com/zapier/zapier-mcp ``` Followed by authentication: ```bash /mcp auth zapier ``` Kiro connects via kiro.dev/powers without a separate installation step. The plugin is also listed on Smithery (smithery.ai/servers/zapier) and PulseMCP (pulsemcp.com/servers/zapier) for users who discover MCP servers through those registries. ## What the Plugin Layer Provides Beyond the Raw Connection A raw MCP server connection gives an AI client the ability to call Zapier tools, but it provides no context about which tools to use first or how to configure them for common use cases. The zapier/marketplace plugin adds three things on top of the base connection. First, guided onboarding walks a new user through connecting their first app and triggering a test action. Second, a quick demo illustrates how the AI client can use Zapier tools in a realistic scenario. Third, role-tailored skills adjust the assistant's behavior based on what kind of work the user describes doing, such as operations, sales, or engineering, so the tool suggestions and workflows reflect common patterns for that role. The README notes that these additions are part of the plugin distributed from this repository and are not part of the hosted MCP server itself. A client that connects to mcp.zapier.com directly without the plugin will have full tool access but without the onboarding layer. For contributors to the plugin, the repository includes AGENTS.md as a guide for AI agents working in the codebase, CONTRIBUTING.md for human contributors, and llms.txt as an LLM discovery index. ## Zapier MCP vs. n8n: Different Approaches to AI-Connected Automation n8n is an open-source workflow automation tool that also has MCP integration. The two tools differ in architecture and deployment model. n8n is self-hosted by default, which means all workflow execution and credential storage happen on infrastructure the user controls. Zapier MCP routes through Zapier's infrastructure, which is cloud-hosted and SOC 2 Type II certified. For teams with strict data residency requirements, n8n's self-hosted model gives direct control over where data flows. For teams that want a managed service with no infrastructure to operate, Zapier MCP requires no server to run. The README describes Zapier MCP as requiring no code and no infrastructure from the user, which is the core operational trade-off compared to self-hosted tools. Zapier's app library is broader in coverage than most self-hosted workflow tools, which can simplify setup for teams that rely on many third-party SaaS tools. The right choice between them depends on the organization's posture toward cloud data routing and how much operational work they are willing to take on. ## What the Repository Does and Does Not Contain The zapier-mcp repository is a thin distribution layer. The top-level entries are: AGENTS.md, CLAUDE.md, CONTRIBUTING.md, LICENSE, README.md, assets/, gemini-extension.json, llms.txt, plugins/, and zapier-power/. The plugins/zapier/ directory contains the plugin code. The gemini-extension.json enables the Gemini CLI installation path. The llms.txt file is an LLM discovery index that helps automated tools find and understand the repository's purpose. The actual MCP server, its tool definitions, the 9,000+ app integrations, authentication, and permission management are all part of Zapier's hosted infrastructure and are not in this repository. Developers who want to inspect or modify the server-side implementation cannot do so from this codebase. The repository's AGENTS.md file provides a guide specifically for AI agents working in the codebase, which is a practical acknowledgment that AI coding tools are among its likely contributors. The repository is MIT-licensed. The last push was on 2026-07-29 and there are no GitHub releases. Changes to the plugin are tracked through commits to the main branch. The zapier-power/ directory is listed at the top level but the README does not describe its contents, so its purpose is unclear from the documentation alone. ## Conclusion Zapier MCP is a practical choice for teams using Claude Code, Cursor, Codex, GitHub Copilot CLI, or VS Code who want their AI client to trigger and query Zapier-connected applications without writing custom integrations. The base connection requires only a server URL and works in any MCP-compatible client. The plugin layer adds a better first-run experience in supported clients. The right question before adopting it is whether Zapier's permission model and data routing match your organization's compliance requirements, since all app interactions pass through Zapier's infrastructure. The last push was on 2026-07-29. ## FAQ ### What is MCP in Zapier? MCP stands for Model Context Protocol, an open standard for connecting AI clients to external tools. Zapier MCP is a hosted MCP server that exposes Zapier's automation capabilities as callable tools, letting any MCP-compatible AI client send messages, retrieve data, and trigger workflows across 9,000+ connected applications in plain English. ### What is the Zapier MCP server URL? The Zapier MCP server is hosted at https://mcp.zapier.com/api/v1/connect with type: http. Clients like VS Code connect directly using this URL. Other clients install the Zapier plugin through their respective marketplaces, which configures the server URL automatically. ### How do you use Zapier MCP with Claude Code? For Claude Code, add the Zapier plugin from the zapier/marketplace or the anthropics/claude-plugins-official repository. This installs the plugin that adds guided onboarding, a first-use demo, and role-tailored skills on top of the base Zapier MCP server connection. ### What does the Zapier MCP plugin add that the raw server connection does not provide? The plugin adds guided onboarding that walks a new user through connecting their first app, a quick demo showing how to use Zapier tools in practice, and role-tailored skills that adjust tool suggestions based on the user's described role. These are absent when connecting directly to the server URL without the plugin. ## Sources - [Issues](https://github.com/zapier/zapier-mcp/issues) - [License: MIT](https://github.com/zapier/zapier-mcp/blob/main/LICENSE) - [Project website](https://docs.zapier.com/mcp) - [README](https://github.com/zapier/zapier-mcp/blob/main/README.md) - [zapier/zapier-mcp on GitHub](https://github.com/zapier/zapier-mcp) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/zapier-zapier-mcp