Discover what you can automate with Zapier by searching its app directory and finding popular ready-made Zap templates.
Search 9,000+ Zapier apps by name or browse most popular apps, returning slugs for template lookup.
Find popular Zap templates for one app or a pair (up to 3 app slugs), each with a title, step chain, description, and one-click create URL.
Get real automation ideas ranked by actual usage instead of hallucinated suggestions.
Use MCP tools search_apps and find_zap_templates; requires a Zapier Partner API client_id (not secret) and optionally ZAPIER_API_BASE.
Provides tools for discovering Zapier automations, searching the app directory, and finding popular Zap templates with one-click create links.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@zapier-discovery-mcp What do people automate between Slack and Notion?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP server for automation discovery. Zapier MCP executes automations — this server answers the question that comes first: what could I automate?
"What do people automate between Slack and Notion?" — real, popular Zap templates, each with a one-click create link
"Is there a Zapier integration for this tool?" — search the 9,000+ app directory
"Give me automation ideas for my CRM" — templates ranked by actual usage, not hallucinated
Sibling project: zapier-dev-mcp for building Zapier integrations.
You need a Zapier Partner API client_id — free with any Zapier integration, and not a secret (it ships in client-side embed code). See the Partner API docs for where to find yours.
Claude Code
claude mcp add zapier-discovery --env ZAPIER_CLIENT_ID=your_client_id -- npx -y zapier-discovery-mcpClaude Desktop — add to claude_desktop_config.json:
{
"mcpServers": {
"zapier-discovery": {
"command": "npx",
"args": ["-y", "zapier-discovery-mcp"],
"env": { "ZAPIER_CLIENT_ID": "your_client_id" }
}
}
}Optional: ZAPIER_API_BASE overrides the API host (testing, proxies).
Related MCP server: mcp-server-mcpindex
Tool | What it does |
| Search the app directory by name; popularity-ordered without a query. Returns slugs for the next step. |
| Popular ready-made Zap templates for one app or a pair — title, step chain ( |
Defensive parsing. Partner API responses are only partially documented; the client whitelists fields and handles both bare-array and enveloped responses rather than assuming shapes.
Testable without credentials. The smoke test runs a local HTTP mock of the Partner API and points the server at it via ZAPIER_API_BASE — full end-to-end over stdio with zero secrets. Missing ZAPIER_CLIENT_ID produces setup instructions as a tool error, not a crash.
HTML in titles/descriptions is stripped and truncated before it reaches the model.
Not affiliated with or endorsed by Zapier.
npm install
npm test # offline unit tests (mocked fetch)
npm run build # tsc → dist/
node scripts/smoke.mjs # end-to-end against a local Partner API mockMIT
What do people actually automate with these apps? Returns popular, ready-made Zap templates (by app slug from search_apps — one app or a pair like ['slack','notion']), each with a step chain and a one-click create link.
Inputs: apps (required), limit.
Search Zapier's directory of 9,000+ integrated apps by name. Without a query, returns the most popular apps. Use the returned slugs with find_zap_templates.
Inputs: limit, query.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 2 tools
The two tools have clearly distinct purposes: one searches/returns apps, the other retrieves Zap templates for a given app or app pair. There is no overlap or ambiguity in what each tool does.
Both tool names follow a consistent verb_noun pattern: search_apps and find_zap_templates. The naming is predictable and aligned with their respective actions.
With only two tools, the server feels minimal but arguably well-scoped for its stated discovery purpose. It is exactly at the borderline between 'too few' and 'appropriately lean' for the domain.
The core discovery workflow (search apps, then find templates for those apps) is covered, but there are noticeable gaps such as searching templates by keyword/use case or getting detailed app information. Agents can complete the basic flow but may hit dead ends for broader discovery scenarios.
Hosted MCP server connecting AI assistants to 9,000+ apps and 40,000+ actions via Zapier.
Build and run automations across 2,300+ apps and build app integrations from any MCP client.
Connect to 1,400+ apps and 15,000+ actions through one OAuth-protected MCP server.
Connect any AI agent to 1,000+ apps and 27,000+ actions through one remote MCP server (OAuth).
