Build your own MCP agent with pre-made prompt-pack
This tutorial guides you step-by-step through connecting Anthropic’s Claude with Zapier MCP, enabling you to automate tasks across hundreds of apps directly from natural-language prompts. This agentic integration empowers you to leverage Claude’s advanced reasoning capabilities to streamline workflows, increase productivity, and easily orchestrate complex actions without coding.
What about ChatGPT’s new agent? Until that’s available via API (currently rolling out to users on desktop), this will be the best way to connect as many tools from your stack with an LLM (via Zapier) versus the ones OpenAI has integrated with.
Let’s go…
Will this save me time? This will likely save you at least 3-5 seconds for each simple action, like asking, ‘What is the latest Slack message from Sarah?’, since you can query your requests directly in natural language instead of navigating your tool’s interface. That can save up to half an hour per task if you’re doing more in-depth data work and analysis, such as summarizing all your open CRM deals and suggesting next steps.
Pre-requisites:
Step-by-Step: Connect Anthropic (Claude) with Zapier MCP
Step 1. Create a New MCP Server in Zapier (2-3 mins)
Log in to mcp.zapier.com.
Click ”+ New MCP Server”.
Select “Claude (Anthropic API client)” from the dropdown.
Name your server (e.g., “Claude–Zapier MCP”).
Click “Create”.
Step 2. Add Tools (Zapier Actions) to Your MCP Server (3-5 mins)
Go to the Configure tab on your MCP server.
Click ”+ Add tool”.
Search/select apps like Slack, Gmail, Airtable, etc.
Choose the specific actions or select “Add all” for convenience.
Connect and authenticate your app accounts (guided by Zapier).
Step 3. Copy Your MCP Server URL (30 seconds)
Go to the Connect tab.
Click “Copy URL”. (This URL connects Claude and Zapier.)
Step 4. Connect MCP Server to Claude (2 mins)
Open Claude (web or desktop).
Navigate to Settings → Integrations.
Click “Add connector”.
Paste the MCP URL copied from Zapier.
Name your connector (e.g., “Zapier MCP”) and click “Add”.
Accept any security prompts.
Step 5. Enable & Test Zapier Actions in Claude (2-3 mins)
Open a new conversation in Claude.
Click Search & Tools (🔍).
Enable your new Zapier MCP connector.
Select which actions you want Claude to access.
Step 6. Run a Test Prompt (1-2 mins)
Enter a natural-language prompt to test, e.g. “Summarize unread emails from today and send the summary to Slack #general.”
Claude will confirm actions and execute tasks via Zapier integrations.
Use the prompt-pack below to try out some further use-cases!
Prompt-pack
Here’s some quick examples to get ideas flowing.
1. Slack
(Team comms, async updates)
“Send a message to the #weekly-review channel: ‘Metrics report is ready in Sheets – check column D for the CTR delta.’”
“DM @Alex: ‘Quick Q—are you still runninguser migration rollout this week?’”
“Post a poll in #product: ‘Which design concept do you prefer?’ Options: Concept A, Concept B.”
2. Google Sheets
(Ops dashboards, reporting)
“Update the ‘Sales Funnel’ sheet: Row for 16 July, Stage 2 conversion: 43%.”
“Create a new tab titled ‘User Feedback – July’ and paste today’s survey data from Notion.”
“Fetch the last 7 days of demo signups and calculate the average time-to-close.”
3. Notion
(Docs, team wikis, internal planning)
“Create a new page in the ‘Go-to-Market’ workspace titled ‘Q3 Launch Timeline’ with an empty table.”
“Summarise today’s meeting notes and append them to the ‘Weekly Leadership Sync’ page.”
“Search the ‘Customer Research’ database for all entries tagged ‘Churn Risk’ and list them.”
4. Gmail
(Outreach, comms)
“Draft an email to jessica@someaicompany.io: ‘Following up on your interest—attached is our updated demo video.’”
“Reply to Raj’s latest thread and confirm we’ll share next steps by Thursday.”
“Forward today’s investor update to my co-founder and ask for feedback before sending.”
5. Google Calendar
(Scheduling, context capture)
“Schedule a 1:1 with Ben for Thursday at 3pm titled ‘UX Audit – Mobile Checkout’.”
“Block 90 minutes this Friday morning for deep work: ‘Q3 Planning – No Meetings’.”
“Cancel tomorrow’s investor sync and offer a reschedule link in the follow-up email.”
6. Airtable
(CRM, pipelines, content tracking)
“Add a new contact: Name: Amy Tran, Company: Continuum, Status: Lead, Notes: ‘Spoke at SaaS’.”
“Update the ‘VC Pipeline’ base: mark ‘Lighthouse Ventures’ as Term Sheet Signed.”
“Search the ‘Beta Signups’ base for users who churned in <30 days and export to CSV.”
7. ClickUp (or Asana)
(Project mgmt, execution)
“Create a task: ‘Review onboarding flow’ → due Friday, assignee: Jay, priority: High.”
“Update the status of ‘Redesign homepage hero’ to ‘In QA’ and move to ‘Sprint 7’.”
“Get a list of overdue tasks from the Growth project and post in Slack.”
8. Google Drive
(Assets, docs, collaboration)
“Upload ‘Team Offsite Deck.pdf’ to ‘Internal → People Ops’ folder and share with HR team.”
“Find the most recent ‘OKRs’ doc in Drive and summarize the key Q3 targets.”
“Create a new folder in ‘Clients → Lumina AI’ and grant edit access to freddie@acompany.ai.”
9. HubSpot (or similar CRM)
(Sales ops, growth)
“Create a deal: Name: ‘Nova Labs Enterprise’, Value: £180,000, Close date: Aug 30, Owner: Samir.”
“Update contact John Li to ‘Re-engaged’ and add the tag ‘Webinar’.”
“Log a call outcome for Jessica Park: ‘Demo complete. Requested trial access.’”
10. Linear
(Product management, issue tracking)
“Create a new issue in ‘Mobile App’ project: Title: ‘Push notifications not triggering on iOS 17’.”
“Move ‘Simplify onboarding copy’ to ‘In Progress’ and assign to @lucia.”
“List all bugs closed in the last 7 days from the ‘Infra’ team and group by priority.”
What about an actual multi-step workflow?
Example 1: “Investor Update Prep”
Audience: Founder preparing an investor update
Tools: Notion → Google Sheets → Gmail → Slack
Prompt to LLM (Anthropic): “Help me prepare this month’s investor update. Pull the metrics from my Notion dashboard, format them in Sheets, and draft an email to my investor list with a summary.”
LLM-powered flow via Zapier/MCP:
Fetch from Notion: Pull KPIs (MRR, churn, CAC) from “Monthly Metrics” database (Notion) for current month.
Update Google Sheets: Add a new row in “Investor Metrics” sheet with current date and fetched KPIs.
Summarise for Email (LLM): Summarise key changes vs last month (e.g., “MRR up 14%, CAC down 8%”).
Send Gmail Draft:
Draft email:
Subject: “July Update – Continued Growth 🚀”
Body: “Hi [Name], quick update on our progress this month… [LLM summary].”Slack Notification: DM yourself: “Investor update draft is ready in Gmail. Review before EOD.”
Example 2: “Bug Escalation → Resolution Flow”
Audience: Product Manager or Engineer
Tools: Slack → Linear → Notion → Google Calendar
Prompt to LLM (Anthropic):
“Log the urgent Slack bug into Linear, notify the team, and schedule a fix session for tomorrow.”
LLM-powered flow via Zapier/MCP:
Trigger from Slack Message: Detect a Slack message tagged with :bug: in #support-bugs channel.
Log Issue in Linear:
Create issue:
Title: “Slack auth bug affecting mobile users”
Description: Include original Slack message + timestamp
Priority: High, Status: Open, Assigned to: engineering@yourcompany.comAppend to Notion QA Tracker: Add entry in “Bug Tracker” db: status=Open, date=Today, tags=Slack, mobile
Schedule Fix Session in Calendar:
Create event for next day 10:00–11:00:
Title: “Bug Triage – Slack Auth”
Description: Pull in Linear issue summary
Invite: dev team
Confirmation to Slack: Post in #engineering: “New Slack auth bug logged in Linear + triage session scheduled for 10am tomorrow.”