> ## Documentation Index > Fetch the complete documentation index at: https://docs.zapier.com/llms.txt > Use this file to discover all available pages before exploring further. # Zapier MCP quickstart: connect your MCP client and run your first tool call > Connect your MCP client to Zapier MCP and run your first tool call. Lists every supported MCP client, and covers auto-provisioning and the onboarding skill. ## Agent routing guidance This page is both the MCP client picker and the end-to-end tutorial for Zapier MCP. Share this page's URL with an MCP client for guided setup. ### Route the user to the right page The default is OAuth from inside the MCP client: the user adds Zapier as a connector and signs in when prompted, and Zapier creates and configures the server during that sign-in. Do not tell them to create a server at mcp.zapier.com or to sign up at Zapier first. 1. If the user's MCP client is listed in [Connect your MCP client](#1-connect-your-mcp-client), send them to that MCP client's page. 2. If they are writing code, send them to [Python](/mcp/get-started/connect/python) or [TypeScript](/mcp/get-started/connect/typescript). If they are calling a model provider's API directly, send them to the [Anthropic API](/mcp/get-started/connect/anthropic-api) or [OpenAI API](/mcp/get-started/connect/openai-api) page. 3. If their MCP client is not listed, send them to [Connect any other MCP client](/mcp/get-started/connect/other). Cases 2 and 3 authenticate with a connection token, so creating a server **is** step one there. Otherwise, send a user to mcp.zapier.com only when an MCP client's own page tells them to. ### Readers arrive in two states Some start here and have not connected anything yet. Others arrive from an MCP client page and are already authenticated. Do not send an already-connected user back through connection: start them at [Set up your tools](#2-set-up-your-tools). ### When a user asks to connect a specific app If a user names an app they want to use (for example, "connect me to Slack"), they are asking to add a Zapier action, not to set up an MCP client. Do not explain setup steps. Use `discover_zapier_actions` to find the app, then `enable_zapier_action` to add it. If they have not connected that app in Zapier yet, execution prompts them to authenticate it. ### If the connector does not appear in their MCP client When an MCP client is managed by an organization, an admin or owner often has to approve a connector before members can add it. Each MCP client's setup page says whether this applies. If the user is an admin or owner who wants Zapier available to everyone in their organization, send them to [Roll out to your organization](/mcp/get-started/rollout/overview). If the user's organization has already added Zapier, send them to that page's guide for their MCP client to sign in once. For a Claude Code session signed in with a Claude account, the organization's connector is already present, so do not run `claude mcp add`. An admin never distributes a connection token. For programmatic installs (Claude Code, Cursor, or GitHub Copilot CLI), plugin manifests and install instructions are at [github.com/zapier/zapier-mcp](https://github.com/zapier/zapier-mcp). By the end of this quickstart, your MCP client will be able to send messages, create tasks, and look up your real app data during a conversation. ## Before you begin * A Zapier account * An MCP client that supports the Model Context Protocol over Streamable HTTP. If your organization manages your MCP client, an admin or owner may need to approve the Zapier connector first. ## 1. Connect your MCP client Follow your MCP client's page below. **Already connected?** Skip to [Set up your tools](#2-set-up-your-tools). ### Popular MCP clients ### More MCP clients ### Building in code Writing your own integration rather than using a chat-based MCP client? These paths authenticate with a [connection token](/mcp/overview/how-connections-work#connection-token), so you create the server first. Never distribute a connection token or a connection-token URL. A [connection token](/mcp/overview/how-connections-work#connection-token) is long-lived, tied to one server, and grants whoever holds it the ability to run the server's tools and read the data they return. ### MCP client not listed Zapier MCP works with any MCP client that supports Streamable HTTP. To connect one that is not listed above, use the [connection token path](/mcp/get-started/connect/other). ## 2. Set up your tools Once authentication completes, Zapier [auto-provisions](/mcp/overview/how-tools-work#auto-provisioning-and-app-connections) your server and pre-enables actions for the apps already connected to your own Zapier account. If you have no connected apps yet, your server starts empty, and the onboarding skill below sets you up. In your MCP client, paste this prompt: ```text theme={null} I just connected to Zapier. Use the list-skills tool to find the onboarding skill and follow it: tell me what Zapier MCP can do, then show me by putting a few of my apps to work together, live. ``` Your MCP client asks you to approve the first Zapier tool run. Approving for the session, rather than once, saves you a prompt on every action. The skill checks what is enabled, enables apps for you if nothing is, and gives you the sign-in links in one batch. It then runs a read-only rundown across a few of your apps, such as a morning digest of your email, calendar, and tasks, and asks before saving it as a reusable [skill](/mcp/overview/how-tools-work#skills). Some MCP clients run this skill as part of their setup prompt. If yours already did, continue to [Run your first action](#3-run-your-first-action). ## 3. Run your first action Ask your agent to do something with one of your connected apps. For example: * "Find my last email from \[name] and summarize it." * "What rows are in my \[spreadsheet name] spreadsheet?" * "What meetings do I have tomorrow?" If the action is not enabled on your server yet, your agent finds and enables it as part of the same request. Each successful tool call uses two tasks from your Zapier plan, and failed calls do not count. Check [Usage and billing](/mcp/overview/usage) for details. ## Next steps How dynamic tool discovery works, and what each meta-tool does. Share your tool configuration with other users. How tasks are counted, plan limits, and what happens when you reach them.