Connectors & Workflows
AI that does things,
not just answers.
A prompt is a conversation. A workflow is a machine. Here's how to connect AI to your calendar, email, spreadsheets, CRM, and everything else — so it actually acts on your behalf.
The concept
From prompt to pipeline.
When you type a question into ChatGPT or Claude, you're having a one-off conversation. Connectors change this — they create persistent, automated pipelines where AI listens for triggers, processes information, and takes action across multiple tools without you having to ask each time.
01
Trigger — something happens
A new email arrives. A form is submitted. A spreadsheet is updated. A calendar event is created. The workflow starts when a defined event occurs — automatically, without you needing to initiate anything.
02
AI processes — it thinks about what to do
The content is sent to an AI model (ChatGPT, Claude, Gemini, etc.) with a defined prompt. The AI classifies, summarises, extracts, drafts, scores, or analyses — depending on what you've instructed it to do.
03
Action — something gets done
The AI's output triggers an action: a reply is sent, a row is added to a spreadsheet, a Slack message is posted, a task is created in your project tool, a document is generated. Work gets done — without you in the loop.
Example workflow — Email triage
Inbound email → AI classifies → routed to right person
📧 New email arrives
→
🤖 AI reads & classifies
→
📋 Tags: "Sales enquiry"
→
📤 Forwards to sales team
→
✅ Draft reply created
Example workflow — Content pipeline
Meeting transcript → AI summary → distributed to team
🎙️ Meeting ends
→
📝 Auto-transcribed
→
🤖 AI extracts actions
→
✅ Tasks created in Asana
→
💬 Summary posted to Slack
The tools — updated 2026
The major connectors explained.
Six tools. Different power levels, different audiences. Pick the one that matches where you are — you can always move up.
⚡
Zapier
No-code · 8,000+ apps
Beginner-friendly
The most accessible automation platform. Connects over 8,000 apps with a simple trigger-action model. In 2026, Zapier's AI Steps are significantly better — you can chain multiple Claude or ChatGPT steps in sequence, with each step using the output of the last. If you can describe what you want in plain English, you can build it in Zapier.
Best for
Non-technical users who want quick wins
Simple 2–4 step automations with AI in the middle
Connecting mainstream business tools fast
Email triage, lead routing, content pipelines
Free: 100 tasks/month · Starter: $19.99/month
🔄
Make
Visual · Advanced logic
Intermediate
Formerly Integromat. A visual workflow builder that handles complex, multi-step scenarios Zapier can't. Flowchart interface lets you branch logic, loop over datasets, and handle errors gracefully. A major upgrade in 2026: Make's AI module now supports Claude natively with structured output — meaning AI responses can feed cleanly into downstream steps without parsing.
Best for
Complex multi-step workflows with branching logic
Looping through lists (emails, rows, files)
European users — strong GDPR compliance
Better value than Zapier at higher volumes
Free: 1,000 ops/month · Core: $9/month
🔀
n8n
Open-source · Self-hostable · AI agents
Developer-friendly
The most powerful and flexible — and the most technical. Open source, self-hostable. In 2026, n8n's AI agent nodes are mature enough to build production-grade agentic pipelines: multi-step reasoning, tool use, memory, and sub-agents that spin up other agents. The go-to platform for anyone building real AI automation systems rather than simple zaps.
Best for
Building production AI agents with memory and tools
Self-hosting for full data control and privacy
Custom JavaScript/Python inside workflows
High-volume automations — no per-task pricing
Self-hosted: Free · Cloud: $20/month
🤖
Lindy
AI-native agents · No-code · 2026
Beginner-friendly
Lindy is the new breed of automation tool — built AI-first, rather than retrofitted. Instead of building workflows step by step, you create "Lindies" — AI agents you describe in natural language. Tell it what you need, and it figures out the steps. Particularly strong at email management, meeting prep, lead qualification, and CRM updates. The fastest way from zero to a working AI agent.
Best for
Email triage and intelligent auto-responses
Lead research and CRM enrichment
Meeting prep — briefings before every call
Anyone who finds Zapier too manual
Free tier available · Pro: $49/month
🔁
Relay.app
Human-in-the-loop · Collaborative automation
Intermediate
Relay is built around the idea that not every automation should run fully autonomously — sometimes a human needs to review or approve before the next step fires. It combines Zapier-style automation with human checkpoints: AI drafts the email, a human approves it, then it sends. Perfect for customer-facing workflows where full automation is too risky but full manual is too slow.
Best for
Customer communication that needs human review
Sales outreach with AI drafts, human approval
Compliance-sensitive workflows
Teams that want AI assistance, not AI replacement
Free: limited runs · Pro: $9/month per user
🖥️
Claude Desktop + MCP
Native · No middleware · Direct tool access
Intermediate
Claude's desktop app now connects to your tools directly via MCP — no Zapier, no middleware. Install an MCP server for GitHub, Slack, Notion, or your database, and Claude can read from and write to those tools inside the conversation. "Check my latest 10 GitHub issues." "Summarise what was decided in today's Slack threads." Real actions, real data, no copy-paste.
Best for
Developers and power users who want direct tool access
Working with real data inside Claude conversations
No per-task pricing — just your Claude subscription
Privacy — data stays local, not routed through third parties
Included with Claude Pro/Team · MCP servers free at mcp.so
2026 workflow patterns
What people are actually automating now.
These are the workflows that have moved from "interesting experiment" to "running in production" in 2026. Each one is achievable with the tools above.
Agentic lead qualification
Inbound lead → researched, scored, and routed before a human looks
New form submission → AI researches company (web search) → scores against ICP → drafts personalised intro email → routes to correct sales rep with a briefing → logs everything to CRM. Zero human time until the rep opens their inbox.
📬 Form submitted
→
🔍 AI researches company
→
📊 ICP score: 82/100
→
✉️ Draft email created
→
📋 CRM updated + rep notified
BUILD WITH: Lindy or n8n + Claude API
Deep research pipeline
Topic brief → multi-source research → structured report, daily
Every morning at 7am: AI runs web searches on defined topics, reads full articles, cross-references sources, writes a structured research brief with key findings and source links — delivered to your inbox before you start work. One setup, daily intelligence, zero effort.
⏰ 7am trigger
→
🌐 AI searches 8 sources
→
📖 Reads & cross-references
→
📝 Research brief written
→
📧 In your inbox by 7:30
BUILD WITH: Make or n8n + Claude Deep Research API
Content repurposing engine
One long-form piece → 8 different formats, automatically
Upload a podcast transcript or blog post → AI extracts key insights → writes LinkedIn post, Twitter thread, email newsletter, short-form video script, and Instagram caption — all formatted for each platform. What used to take a content team half a day now runs in 4 minutes.
🎙️ Podcast transcript
→
🤖 AI extracts themes
→
✍️ 8 formats drafted
→
👁️ Human review (Relay)
→
📱 Scheduled to publish
BUILD WITH: Zapier + Relay.app + Claude API
Meeting intelligence
Meeting ends → notes, actions, follow-ups — done before anyone leaves the room
Meeting transcript sent to AI → extracts decisions, action items, open questions, and owners → creates tasks in your project tool → sends personalised follow-up emails to each participant → posts summary to Slack. Your meeting admin is handled before the Zoom window closes.
🎤 Transcript ready
→
🤖 Decisions + actions extracted
→
✅ Tasks in Notion/Asana
→
📧 Follow-ups sent
→
💬 Slack summary posted
BUILD WITH: n8n or Make + Claude API + MCP
The new standard
MCP — the universal connector protocol.
Model Context Protocol is an open standard, developed by Anthropic, that's becoming the universal language for connecting AI models to tools and data — regardless of which AI you use.
What MCP actually does
Before MCP, every AI tool needed its own separate integration with every other tool. Claude needed a Slack integration. ChatGPT needed its own Slack integration. Cursor needed its own. Each one was custom-built and didn't work with the others.
MCP solves this by creating a single standard. Build an MCP server for Slack once — and now Claude, ChatGPT, Gemini, and Cursor can all connect to it without any extra work. One integration, all models.
MODEL-AGNOSTIC
Works with Claude, ChatGPT, Gemini, Cursor — any AI that supports MCP
BUILD ONCE
Create one integration and use it across every AI tool in your stack
GROWING FAST
Thousands of MCP servers already exist for Slack, GitHub, Google Drive, databases, and more
OPEN STANDARD
Not owned by any single company — open source, community-driven, and free to use
What you can connect with MCP
📁
File systems
Read, write, and organise local files and cloud storage (Google Drive, Dropbox, OneDrive)
💬
Communication
Slack, Teams, Gmail — read messages, send replies, summarise threads
📊
Databases
Query Postgres, MySQL, SQLite — ask questions in plain English, get SQL results
🔧
Dev tools
GitHub, GitLab, Jira — read issues, create PRs, manage code reviews through AI
🌐
Web browsing
Let AI browse the web, fill forms, extract data, and navigate interfaces on your behalf
📅
Calendars & tasks
Google Calendar, Notion, Asana — schedule meetings, create tasks, manage projects
Decision guide
Which connector should you use?
🙋
"I just want to automate something and I'm not technical"
Start with Zapier. It's the easiest, has the most tutorials, and works for 90% of common automations. Build your first Zap in 15 minutes.
⚙️
"I need more control and my automations are getting complex"
Move to Make. Visual, powerful, better value for complex workflows. Good for branching logic and looping through large datasets.
🧑💻
"I'm a developer who wants to build real AI agents"
Use n8n (self-hosted). Full code access, 70 AI nodes, LangChain integration, and you own the data. The best platform for serious AI workflow engineering.
🔌
"I want to connect AI to my own tools and don't want to be locked into one platform"
Learn MCP. Build or install MCP servers for the tools you use, and your whole AI stack gets connected at once — independent of which AI model you prefer.
Getting started
Your first automation in 15 minutes.
The fastest way to understand connectors is to build one. Here's the simplest workflow you can make right now — no code, no cost, five minutes.
STEP 01
Create a free Zapier account
Go to zapier.com and sign up free. No credit card needed for the free tier.
STEP 02
Pick a trigger you'll recognise
Start with Gmail → "New email matching search" as your trigger. Use a search term that identifies a specific type of email you get regularly.
STEP 03
Add an AI step
Add "ChatGPT" or "Claude" as the action. Prompt it to summarise the email in 3 bullet points, then extract any action items.
STEP 04
Send the result somewhere useful
Add a final action — save the AI summary to a Google Sheet, post it to a Slack channel, or append it to a Notion page. Turn on the Zap. You're done.