Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
30 contributions to AI Agency Plus +
Self-hosting n8n for clients? the license cares about who can log in
most of us here build on n8n, so i finally read the license FAQ properly. self-hosted n8n runs under their Sustainable Use License, and as i read the FAQ, building and maintaining workflows for clients is explicitly fine. you can charge for builds, maintenance, training, consulting. where it gets tighter is client access. the FAQ says agency work is ok as long as clients don't access your n8n instance directly, and it lists hosting n8n as a service where clients build their own workflows as not allowed. same for white-labelling it. the way they frame it is who controls the workflow logic. two caveats. this is only the self-hosted license, n8n Cloud has its own terms. and some setups are called gray in the FAQ itself, they say to email license@n8n.io to confirm. i'm not a lawyer, this is just my read of their page. FAQ: https://docs.n8n.io/n8n-community-license/community-license/license-faq curious how you all set it up. does each client get their own instance on their own server, or do you run everything on yours and they just get the results?
1 like • 6d
good to know the line is basically who can log in. what about giving a client a view only login just so they can see their executions? feels like that might already be the gray zone they mention
1 like • 6d
@Andrew Moore yeah fair, guessing on licensing is a bad idea lol. not something I need right now but if I do I'll email them and post what they say
Putting an AI bot on a client's WhatsApp? there's a rule about what kind of bot it can be
since a few of us are building on WhatsApp (Amanda's coexistence post got me reading), this one seems worth knowing. TechCrunch reported that WhatsApp changed its Business API terms to bar general purpose AI chatbots. new accounts got the terms in Oct 2025 and it applied to everyone from Jan 15 2026. the way they describe it, the rule is aimed at AI providers where the assistant itself is the main thing being offered, think a ChatGPT style bot living inside WhatsApp. Meta's line in that piece is that the API is meant for businesses doing customer support and sending relevant updates. so the normal agency build, a bot that answers questions about the client's own business, books appointments, sends reminders, looks like it's on the allowed side. respond.io reads it the same way, though they sell WhatsApp bots so keep that in mind. where i'd be careful is the "ask me anything" bot with a client logo on it. if it happily writes essays or answers random trivia, it starts to look like the thing the rule is about. keeping the prompt scoped to the business seems like the safe move anyway. i haven't read Meta's actual terms line by line, only the coverage, so check the wording yourself before promising a client anything. https://techcrunch.com/2025/10/18/whatssapp-changes-its-terms-to-bar-general-purpose-chatbots-from-its-platform/ https://respond.io/blog/whatsapp-general-purpose-chatbots-ban do you lock your WhatsApp bots to business topics only, or let them answer whatever people ask?
1 like • 6d
makes sense, a bot that only answers about the client's own business is basically support anyway. I'd still keep a handoff to a real person in there, feels like the safer side of that line too
0 likes • 6d
@Andrew Moore haven't built one yet so no real answer, but I'd want both. asking for a person is the easy one. the no answer case feels more important though, that's where it would otherwise just guess
Adding SMS to a client automation in the US? register the number first (A2P 10DLC)
a lot of the builds in here send texts. review requests, missed call text back, booking reminders. one thing that seems to catch people the first time is A2P 10DLC. short version as i read Twilio's docs: if you send business texts to US phones from a normal 10 digit number, the business has to register a brand and a campaign first. Twilio says unregistered US bound messages from those numbers have been fully blocked since Sept 2023. their support page also says blocked messages come back with error 30034 and you still get charged for them. so the workflow can look perfect in n8n and the texts just never land. per Twilio the registrations get reviewed by outside vetting partners, so it isn't instant. seems worth kicking it off at the start of a project and not the day before launch. i only read Twilio's pages and other SMS providers have their own process, so check yours: Twilio's notice on unregistered 10DLC: https://support.twilio.com/hc/en-us/articles/14910496447771-Shutdown-of-Unregistered-10DLC-Messaging Twilio's overview of registration: https://www.twilio.com/en-us/blog/a2p-10dlc-registration for those of you sending SMS for clients, do you register under the client's own account or run it through yours?
1 like • 10d
the 30034 part is sneaky. if I get it right the Twilio node in n8n still shows success because the API accepted the message, and the block only shows up later on the status. so probably worth adding a status callback to these builds so failed texts actually get noticed
1 like • 7d
@Andrew Moore same channel I think, one place to look is easier. I'd just tag them sms so they're easy to filter. haven't actually built the callback part yet though, so ask me again once I have lol
Building cold email or newsletter automations for clients? read Gmail's sender rules first
after the SMS one, same idea for email. a lot of outreach and follow up builds send from the client's own domain, and Gmail has its own rulebook for that. as i read Google's sender guidelines: everyone sending to Gmail accounts needs SPF or DKIM set up on the domain. once a domain sends around 5,000+ messages a day to Gmail it counts as a bulk sender, and then it needs SPF and DKIM both, a DMARC record, and one-click unsubscribe on marketing mail. they also say to keep the spam rate in Postmaster Tools under 0.3%, and their FAQ suggests aiming below 0.1%. most small clients won't hit 5,000 a day, but the basic auth part applies to everyone, and a fresh domain with no DNS records set up seems like an easy way to land in spam on day one. Google's guidelines: https://support.google.com/a/answer/81126 their FAQ: https://support.google.com/a/answer/14229414 i only read Google's pages plus a few vendor write ups, so check the current version before you promise a client anything. question for people doing outreach builds: do you set up the client's DNS records yourself, or hand them a checklist and wait?
1 like • 7d
haven't done an outreach build for anyone yet so no real answer on the DNS part, but the checklist route sounds safer to me. I wouldn't love having login access to a client's domain. didn't know about the 0.3% spam rate line though, good to know before promising anything
Free: the Script Agent Starter Kit from my new video
As promised in the video, here are the 3 files behind the AI agent that writes my YouTube scripts: 1. Brand file template: who you are, who you talk to, your proof line 2. Hook library: 8 hook types with fill-in templates + a hook checklist 3. Voice rules: what Claude learned from 23 hours of Hormozi, 12 rules and a banned words list Page 1 shows the 15-minute setup and the exact prompts I use in Claude. Watch the video: https://youtu.be/MQkPdvVL0h0 Which creator would you train yours on?
1 like • 10d
the banned words list is the part I'd grab first lol. did Claude pull the 12 rules out of the transcripts on its own, or did you write them after reading its notes?
1-10 of 30
Emma Shaw
4
78 points to level up
@emma-shaw-4220
Learning Automation and AI

Active 2d ago
Joined Oct 9, 2026