Zapier MCP server

What Zapier MCP server does

Zapier MCP server gives MCP-compatible AI clients access to Zapier’s app ecosystem, which covers more than 9,000 apps and 40,000 actions. Through the hosted service, an assistant can send messages, retrieve information, and trigger workflows using connected Zapier capabilities. The repository also distributes an optional client plugin that adds onboarding, a first-use demonstration, and role-oriented skills in clients that support it.

The server is also a reference tool for Zapier itself. It exposes documentation search, access to documentation pages and OpenAPI specifications, and a way to report problems with documentation content.

How it works

Clients connect to the hosted MCP endpoint at https://mcp.zapier.com/api/v1/connect using HTTP. The assistant can start with search_zapier for broad or conceptual questions. Search results include contextual text, page titles, and links. To inspect a complete page, the assistant uses the returned path with the documentation filesystem tool and adds the .mdx extension.

query_docs_filesystem_zapier provides a read-only, in-memory filesystem rooted at /. It contains Zapier documentation pages and OpenAPI specifications only; commands do not execute on the user’s computer, the server host, or an external network. The tool supports shell-like inspection and text-processing commands such as tree, ls, rg, cat, head, jq, and grep.

Each filesystem call is stateless: the working directory resets to /, and shell state is not retained between calls. Paths should be discovered with tree, ls, or search rather than guessed.

Setup and configuration

Connect an MCP client with the hosted URL and HTTP transport. VS Code can connect directly using the installation badge or by adding the endpoint manually. Cursor provides a server installation link, and Gemini CLI can install the repository extension with:

gemini extensions install https://github.com/zapier/zapier-mcp

After installation in Gemini CLI, authenticate with /mcp auth zapier. Claude Code, Cursor, OpenAI Codex, GitHub Copilot CLI, Claude Cowork, Kiro, and Gemini CLI can also use the distributed plugin or extension paths described by the project. Some clients receive guided onboarding and role-specific skills through that layer; VS Code connects directly to the hosted server.

Tools and capabilities

• search_zapier searches Zapier’s knowledge base for documentation, code examples, API references, guides, and implementation details.
• query_docs_filesystem_zapier reads and searches the virtual documentation filesystem with read-only commands.
• submit_feedback reports incorrect, outdated, confusing, incomplete, or broken documentation content.
• OpenAPI specifications in the mounted documentation filesystem support research into endpoints, schemas, parameters, and authentication.
• The hosted connection exposes Zapier-connected app actions for tasks such as sending messages, retrieving data, and triggering workflows.

Limitations and notes

The documentation filesystem is not a general shell and cannot modify files, run processes, access a real machine, or make network requests. Documentation paths are site-specific, so broad search or filesystem discovery should precede page reads. The feedback tool is intended for problems with Zapier documentation, not product support requests or feedback about the MCP client itself. The plugin distribution is separate from the hosted MCP connection: clients can connect directly when they support the HTTP endpoint.