I kept seeing MCP everywhere and kinda nodded along without really getting it, so I spent the weekend figuring it out. Sharing the simple version in case anyone else is in the same spot. MCP (Model Context Protocol) is basically a standard plug that lets an AI like Claude talk to your tools. Instead of building a custom integration every time, you connect an MCP server once and the AI can use it. Think USB-C but for AI apps. Why it matters for us: n8n now has an MCP Server Trigger and an MCP Client Tool node. So your n8n workflows can be called by Claude as tools, and your AI agents in n8n can use other MCP servers. 3 ways I think this is actually useful for client work: 1. Turn workflows you already built into tools. Got a "create invoice" or "add lead to CRM" workflow? Expose it with the MCP Server Trigger and the client can just ask Claude to do it in plain English. 2. Internal assistant for the team. Connect their Google Drive, calendar, and CRM through MCP and the team gets one chat where they can ask "what's the status on the Smith job" instead of digging through 4 apps. 3. Keep a human in the loop. Let the AI read and draft freely, but anything that sends, pays, or deletes goes through an approval step. Clients feel way safer and honestly so do I lol. Resources if you want to dig in: What MCP is (official): https://modelcontextprotocol.io n8n MCP Server Trigger docs: https://docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-langchain.mcptrigger/ List of ready made MCP servers: https://github.com/modelcontextprotocol/servers Still learning this myself so if I got something wrong, correct me. Has anyone here shipped an MCP setup for a client yet? Would love to know what it was and if they actually used it

Was reading a few agency breakdowns this week and one point kept showing up: the actual n8n/Make build is a small slice of the work. The rest is figuring out what the client really needs, teaching them to use it, and support after launch. One writeup put it at roughly 20% building and 80% everything else. That's their number, not some official stat, but it matches what I see. Same sources say clients don't care what tool you used. They care about leads, hours saved, or revenue. So pitching "I'll build you an n8n workflow" is weaker than "I'll cut your follow-up time from 2 days to 10 minutes." Two things I'm trying to do better on: 1. Ask what happens when the task gets missed before I talk about any tools 2. 2. Quote the outcome, then the build, then a small monthly support fee If you want to read the stuff I found: https://growwstacks.com/blog/is-n8n-dead-in-2026-automation-truth-after-claude-code-ai-agents https://ideaproof.io/guides/ai-automation-agency Curious what the rest of you do. How do you handle the post-launch part, do you charge for support or just fix stuff when it breaks?

Saw a thread in another group where someone wanted Claude hooked up to WhatsApp but didn't want to lose the normal Business app on their phone (calls, chatting by hand etc). That used to be either/or on one number. It isn't anymore. Meta has a "coexistence" mode now. The client keeps the WhatsApp Business app on their phone and the Cloud API runs on the same number, so your automations and the owner share one inbox. How I'd use it for a local business: 1. Booking confirmations and reminders go out through the API (n8n has a WhatsApp Business Cloud node), the owner still replies by hand from the app 2. After hours auto reply that collects the details and hands it to the owner in the morning 3. Keep a human approval step for anything that isn't a simple template message Stuff to check before you promise it to a client: - the number has to already be on the Business app and the app needs to be up to date. A number that started as API only can't be switched - setup goes through a provider's embedded signup and there's a sync window after onboarding (Telnyx says 24h), so do it on a call with the client - messages sent from the app stay free, API messages follow normal Cloud API pricing Docs I read: 8x8 coexistence guide: https://developer.8x8.com/connect/docs/whatsapp/whatsapp-business-app-coexistence Telnyx coexistence docs: https://developers.telnyx.com/docs/messaging/whatsapp/coexistence Has anyone set this up for a client yet? Curious if the old chat history actually came over properly

Saw this in the n8n release notes and figured it's worth sharing, since a lot of us grab older templates. n8n 2.43.3 (Oct 9) now blocks you from creating or updating a workflow that has deprecated nodes in it. So a template from 2024 or early 2025 might import fine, then refuse to save once you change anything. What I'd do: 1. Before editing, open each node and look for the deprecated warning 2. Swap those for the current version of the node (most have a direct replacement) 3. Save and test with dummy data 4. If you have client workflows on an older self-hosted version, check them BEFORE you update the instance, not after Also in 2.43.5 Gemini models now show by name in the model picker, nice if you're testing cheaper models for simple steps. There's also talk that n8n 3.0 might only support Docker for self-hosting (no more npm). I couldn't confirm that from n8n itself yet, so don't take it as fact, but worth watching if you installed with npm. Release notes: https://github.com/n8n-io/n8n/releases Quick highlights: https://releases.sh/n8n/n8n/highlights Anyone here still running client stuff on npm installs? Curious how you're planning to handle it