Zapier MCP Plugin
Give any MCP-compatible AI client governed access to 9,000+ apps through Zapier's hosted server — send messages, pull data, and trigger workflows in plain English.
- Source repo
- zapier/zapier-mcp
- Stars
- ★ 449
- Last updated
- 2d ago
- License
- MIT
- FA score
- 41/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude CodeClaude.ai (Partial support)
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer using Claude Code who wants their assistant to send messages or operate data in Slack, Gmail, Notion, and similar apps directly
- Not a fit if
- Teams needing integrations with apps outside Zapier's 9,000+ catalog who must build custom connectors
- Organizations that require a fully self-hosted MCP server with no dependence on Zapier's hosted cloud
- Source review
- 41/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
zapier/zapier-mcp is Zapier's official plugin distribution repository, whose core component is the client-side plugin in plugins/zapier/. It connects AI clients such as Claude Code, Cursor, and Gemini CLI to Zapier's hosted MCP server (https://mcp.zapier.com/api/v1/connect), giving them governed access to 9,000+ apps. The plugin itself does not execute automations; it equips your assistant with guided onboarding, a first-use demo, and role-tailored skills so it knows how to use Zapier well. The server side is operated by Zapier, is SOC 2 Type II certified, and requires no infrastructure from the user. The repository also ships one-click install badges for VS Code, marketplace entry points for Claude Code and Codex, and contributor resources including AGENTS.md, CONTRIBUTING.md, and llms.txt.
The repo distributes the plugins/zapier/ plugin, which does three things once installed in an AI client: onboards the user to Zapier MCP, provides a first-use demo, and loads role-tailored skills. The actual app connections are made by the hosted Zapier MCP server: the AI client calls server-side tools over MCP to send messages, pull data, and trigger workflows, all described in natural language. Onboarding only requires a server URL (https://mcp.zapier.com/api/v1/connect, type http). Gemini CLI users install the repo's extension with gemini extensions install and then authenticate with /mcp auth zapier.
- A developer using Claude Code who wants their assistant to send messages or operate data in Slack, Gmail, Notion, and similar apps directly
- An engineer using Cursor who wants to trigger Zapier workflows in natural language during coding sessions
- A Gemini CLI user who wants to install the Zapier extension and authenticate quickly with /mcp auth zapier
- A VS Code user who wants to connect to the hosted MCP server with a one-click install badge
- A team with compliance requirements (SOC 2 Type II) that needs governed AI access to business apps
How do you install or deploy this agent?
Most clients only need the Zapier MCP server URL — see docs.zapier.com/mcp/clients for step-by-step instructions. For clients that install the plugin distributed from this repo:
- Claude Code / OpenAI Codex / GitHub Copilot CLI: add zapier/marketplace or anthropics/claude-plugins-official
- Cursor: use the install badge, or add cursor/mcp-servers
- Claude Cowork: add anthropics/knowledge-work-plugins
- Kiro: via kiro.dev/powers- Gemini CLI:
gemini extensions install https://github.com/zapier/zapier-mcp- VS Code connects directly to the hosted server (not via the plugin): use the badge or manually add https://mcp.zapier.com/api/v1/connect (type: http).
How do you use this agent?
After installing the plugin, ask your AI client in plain English to perform Zapier actions — sending messages, pulling data, or triggering workflows. Gemini CLI users must authenticate after installing the extension:
/mcp auth zapierOn first use, the plugin walks you through guided onboarding and a demo. A Zapier account able to reach the Zapier MCP service is required.
What are this agent's strengths and limitations?
- Covers 9,000+ apps with no code and no infrastructure to run yourself
- The hosted MCP server is SOC 2 Type II certified and provides governed access
- The plugin ships guided onboarding, a first-use demo, and role-tailored skills, lowering the adoption barrier
- Supports Claude Code, Cursor, Codex, Gemini CLI, VS Code and more, and is listed on Smithery and PulseMCP registries
- Core connectivity depends on Zapier's hosted cloud server (mcp.zapier.com) and cannot be fully self-hosted
- The repo contains only the distribution layer; actual capabilities live server-side and cannot be audited from code
- The MIT license covers only the plugin code; using Zapier's service may carry account and fee requirements (pricing is not stated in the source)
- The README does not detail Zapier plan tiers, quotas, or the exact tool list — check docs.zapier.com before committing
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Zapier MCP Plugin This agent | 41 · Major gaps | Agent plugin / skill | ★ 449 | 2d ago | — | Codex · Claude Code |
| Postiz Agent CLI | 54 · Major gaps | CLIFreemium | ★ 544 | 3d ago | TypeScript | Claude Code |
| PipesHub Enterprise AI Context Platform | 66 · Some gaps | Self-hosted serviceFree + model costs | ★ 3.8k | today | Python | — |
| monday.com MCP Server | 60 · Some gaps | Hosted serviceFree + model costs | ★ 436 | today | TypeScript | ChatGPT · Claude.ai |
How does FollowAgents rate this agent?
Why each dimension lost points
Visible evidence is only README and LICENSE; the plugin body (plugins/zapier/) is absent from the evidence set. Least privilege, confirmation, data flow, sensitive data, dependency security, and external effects rest on claims (SOC 2, hosted server) that no source code verifies — 1 each; rollback: 0 (no mention); source attribution clear via MIT + copyright Zapier Inc. — 2.
README is internally consistent but thin on verifiable detail; availability depends on the hosted mcp.zapier.com server, checkable only indirectly via docs links — 1; no error/failure message material — 0.
Audience and scenarios are reasonably clear (AI-client users, guided onboarding) — 2; environment fit covers many clients with concrete install paths — 2; capability boundaries are vague ('9,000+ apps') — 1; trigger precision has no tool/schema detail — 1.
README is well-structured with a repo map and third-party registry listings — 2; install notes per client are concrete — 2; naming stable (zapier/zapier-mcp, plugins/zapier/) — 2; few examples, no FAQ — 1; no known-limitations section — 0; full MIT — 3; versioning depends on external MCP Registry badge, no CHANGELOG — 1; maintenance paths (AGENTS.md, CONTRIBUTING.md, llms.txt) referenced but their contents not provided — 1.
Actual plugin output and tool usability cannot be statically verified — 1; marginal value (guided onboarding, role-tailored skills) is plausibly described — 2; low friction (one server URL) makes cost/benefit reasonable — 2.
Key claims ('9,000+ apps', SOC 2) are not traceable inside the repo, only externally linked — 1; README/badges/registry claims align but lack independent corroboration — 1; marketing language and factual claims are intermixed without separation — 1.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- The plugin body files are not present in the evidence set; all security conclusions rest on README claims and cannot be statically verified.
- Connecting to the hosted server routes data through a third party (Zapier); SOC 2 etc. are external claims to verify independently.
- No known-limitations or rollback guidance exists; review your authorized apps and action scopes in the Zapier console before use.