Zapier MCP: thousands of app actions, with a short allowlist

Zapier MCP connects an MCP client to actions in apps connected to your Zapier account, so an assistant can look up information or make a requested change. For a short, predictable action list, choose managed mode and enable only the actions needed for the task. The official overview describes the service; the setup below treats the action list as an allowlist.

What it can do

Zapier documents app operations such as finding calendar events, adding spreadsheet rows, sending Slack messages and creating Asana tasks. These are examples of available capabilities, not a reason to enable every action. Start with a deliverable you can inspect in the destination app. Zapier MCP overview, as of 30 Oct 2026.

For a support handoff, prepare the message locally, then authorize posting that approved text to a designated internal channel. For a meeting follow-up, approve a task description before allowing its creation in the agreed project. For a spreadsheet lookup, define the sheet and the question before exposing a search action. These are suggested task designs; confirm the action and fields in your account before proceeding.

Write down the destination, permitted change and evidence of completion. A useful acceptance record contains the returned link or identifier and your check of the resulting object. Compare a narrower integration in the Asana MCP guide or Slack MCP guide, and browse MCP servers for apps for other targets.

Keep the list short with managed mode

Zapier distinguishes agentic mode, where the assistant can discover and enable actions, from managed mode, where you choose a fixed toolset and the agent cannot add actions. New servers normally start in agentic mode. How Zapier MCP tools work.

For a bounded pilot, use managed mode. A prompt asking the assistant to stay within a list is useful guidance, but the server configuration should enforce which actions are available. Review the client's other enabled servers too: this allowlist governs its Zapier tools, not every capability the assistant has elsewhere.

Keep separate decisions separate. Looking up a record and updating it have different acceptance criteria. Finish the lookup task, review the result, then authorize a specific update. Read the MCP security checklist before exposing business data.

How to connect it safely

1. Define the first task. Choose a read-only lookup with non-sensitive test data. Record the app account, expected result and stopping point. Prepare any later write as a separate task with approved content and a named destination.

2. Connect the editor client. Zapier's VS Code guide requires MCP support through GitHub Copilot in Agent mode. Use its installation link, or run MCP: Open User Configuration, add Zapier under servers with the URL https://mcp.zapier.com/api/v1/connect, and authenticate when prompted. Follow the official VS Code recipe for the current configuration format. Complete authentication before requesting any app operation.

3. Select managed mode before the pilot. At mcp.zapier.com, select your server, open Settings, choose Managed mode and save changes. This requires Owner or Editor access and changes the mode for every client using that server. Coordinate with other users first. Reload the VS Code window afterward, as described in switching modes.

4. Reduce the enabled actions. Review the server's Apps list and keep only what the current task requires. Inspect existing entries rather than assuming a newly connected server is empty. Confirm the editor's refreshed tool list matches your intended allowlist before authorizing a call.

5. Configure the account and fields. The server owner can use Apps → action menu → Configure. Choose the intended connection. Where the field supports it, set a fixed destination or restrict a dropdown to approved choices; leave generation for content that can vary. Save the configuration. Zapier's action configuration reference explains which field types support each option.

6. Run and verify the bounded task. Approve the specific call in the client, inspect its result and compare it with the source app. For a later write, approve the final destination and payload before execution. Check the created or changed object yourself. Stop once the acceptance condition is met; investigate an ambiguous outcome before authorizing another write.

For editor permission settings beyond this connection, see MCP in VS Code. Recheck the allowlist whenever the intended workflow changes.

What usage means

As of 30 Oct 2026, the published Zapier MCP rate is two tasks per successful external-app tool call. Successful tests count; failed calls, tool discovery and authentication do not. MCP shares your existing Zapier plan allowance, with no separate MCP billing. Usage and billing.

A plain-language request can involve several calls. Estimate the operations you intend to permit, then compare actual usage before expanding the pilot. Budget the editor client's usage separately.

When this does not work

The client uses the wrong transport. Zapier requires Streamable HTTP and does not support SSE. Check compatibility before troubleshooting app permissions. Connection requirements.

The visible list is stale. After changing server mode, refresh the client before relying on its displayed tools. Mode changes can also affect colleagues on the same server. Refresh instructions.

The wrong account or destination is configured. Inspect the action's connection and fixed fields before running it. Field configuration and account-level admin restrictions are separate controls. Configuration boundaries.

The account reaches its allowance. Zapier says calls stop when tasks run out unless applicable pay-per-task billing allows continued use up to a cap, with extra charges. Review those settings before a pilot. Task-limit behavior, as of 30 Oct 2026.

If the proposed operation has no suitable action or configurable field, check the current documentation. Do not substitute a broader permission just to finish the task.

Where Tvijo fits

As of 30 Oct 2026, the DSHREM VS Code extension provides an editor entry to Tvijo tasks, with selected context, task controls and usage receipts. Tvijo's editor guide explains the entry point. TvijoCode is a separate editor test build with the extension embedded; its documented test-build limits include unsigned distribution and no automatic updates.

Tvijo MCP makes Tvijo callable by a compatible client for consultation. Supplied-diff review is prepared, with publication pending in the dated release account; editor integration and safe delegation remain planned work. These components do not establish a ready Zapier integration. A customer-specific connection needs its own scope and acceptance checks.

Use one bounded action per task: approved input, a small action allowlist, a result and verification. A Tvijo task cap does not cap Zapier's task consumption. Read Tvijo costs and budgets separately.

At Tvijo's workspace, as of 30 Oct 2026, Try is $0: up to 5 tasks, 50,000 tokens or $0.05 of sponsored usage, whichever is reached first, from a shared $1 daily pool funded by TVIJO. Service credit is shown as $10 for requesting prepaid credit, with metered provider usage plus an 8% service markup. Payment collection is not activated; an order is not a charge.

FAQ

Is Zapier MCP the same as Zapier AI Actions?

Zapier's official documentation index identifies AI Actions and Natural Language Actions as retired and directs users to MCP. For a new connection, follow the current MCP documentation rather than an older AI Actions tutorial.

Can I prevent the assistant from enabling more actions?

Managed mode provides the fixed action set described in Zapier's tool documentation. Check the mode and the refreshed tool list before starting.

Do successful test calls use Zapier tasks?

Yes. Successful tests count under the same documented rate as other successful app calls. Review usage and billing before testing against connected apps.

Zapier is a trademark of its owner. Tvijo is not affiliated with or endorsed by it.