Recommendations
The tools we actually use
Software and services we run our own business and our clients’ sites on. If we would not put a client on it, it is not on this page.
How to read this page
Some links on this page earn us a commission. If you sign up through them, we may receive a referral fee or credit at no extra cost to you, and in some cases you get a discount. Those links are marked affiliate link so you always know which is which.
The commitment behind every recommendation on this site is simple, and it is the one we are willing to be held to:
- We recommend only what we use ourselves or built ourselves. Every tool on this page runs our own business, this website, or a site we maintain. If we have not run it in production, it is not here.
- Payment does not buy a listing. We do not add a tool because someone offers a commission. The commission, where one exists, follows the recommendation; it never leads it.
- You always know when a link pays us. Affiliate links are labeled at the link, not just in the small print, so you can judge the recommendation with that in hand.
We would rather tell you we earn a fee and have you trust the recommendation than hide the fee and have you doubt everything else on the site. If a tool we recommend stops being one we would use, it comes off this page.
Hosting
Hostinger
Shared hosting and VPS. We run production sites on it, including this one, and it is what we usually recommend for small and mid-sized WordPress sites that need LiteSpeed caching without managed-host pricing. The VPS tier is what we reach for when a site needs Docker, its own database or a staging copy.
Where it fits: WordPress sites up to moderate traffic, and small application servers. Where it does not: sites needing guaranteed support response times or strict compliance hosting.
Visit Hostinger (affiliate link)
What we run on it, specifically
We are not describing this from a feature list. One of their KVM plans runs several of our production WordPress sites behind a Traefik reverse proxy, plus n8n for automation and a Supabase stack, all in Docker on Ubuntu. That is the honest test of a host: not the launch, but what it is like eighteen months in.
What it does well. Full root access and enough headroom to run real container workloads on a plan that costs a fraction of managed hosting. LiteSpeed on the shared tier is genuinely fast for WordPress once its cache is configured properly. Rebuilding a box is quick, and the panel stays out of the way.
What to watch for, from running it. Two things have caught us, and neither is in any brochure:
- No swap is configured by default. On a memory-hungry stack that turns a busy moment into the kernel killing a container instead of slowing down. Add swap, and consider an out-of-memory handler, before you need them.
- A firewall rule does not do what you expect once Docker is involved. Docker publishes ports by writing its own rules, which sit ahead of the usual firewall, so a port you believe is closed can be open to the internet. Ports have to be restricted in the rules Docker respects, and then verified from outside the machine.
- Support is chat-first. Fine for account questions, not a substitute for someone who knows your stack. If you need a guaranteed response time on a production outage, price a managed host instead.
Both of those are configuration, not defects, and both are the kind of thing we set up as a matter of course. They are also why “it is cheap” is the wrong reason to choose a VPS on its own.
The Hostinger name and logo are trademarks of Hostinger International Ltd, used here under its affiliate program. We are an affiliate, not a reseller, partner or certified provider. Plans, specifications and pricing are Hostinger’s and change over time; check their site for current detail.
Automation
n8n, self-hosted
n8n is the workflow engine behind a lot of what we build. We run our own instance on our own server rather than on the vendor’s cloud, which matters for two reasons: the data in those workflows is client data, and pricing that scales per execution punishes exactly the automations that are working.
What we use it for. Moving data between systems that were never designed to talk: webhooks from an ERP into a CRM, scheduled reports that email themselves, order and inventory syncs, form submissions routed to the right person, and the small recurring jobs that otherwise become somebody’s morning routine.
What to plan for, from running it in production.
- A stopped workflow is silent. Nothing alerts you that an automation quietly went inactive after an edit or a restart, and you find out when somebody asks where last week’s report went. Anything that matters needs a heartbeat you actually watch.
- Upgrades can change node behavior. Self-hosting means you choose when to upgrade, which is an advantage right up until you skip six versions. Upgrade deliberately, on a copy first, for anything running live.
- Self-hosting means you own it. Backups, the encryption key that protects stored credentials, and access to the editor are all yours to get right. That is the trade for keeping client data on your own infrastructure.
If you want this without running the server, that is the work: we design the workflows, host them, monitor them and fix them when a vendor changes an API. See CRM consulting for how it connects to the rest of your systems.
Voice, phone and messaging
ElevenLabs
This is our phone system. Calls and SMS run through it, and AI voice agents handle and process calls rather than sitting in a demo. We did not adopt it to write about it; we adopted it because we needed a phone system that could do this, and it is the clearest example we have of the advice on our AI strategy page being something we do to ourselves first.
Where it fits. Businesses that want calls answered, qualified, logged and followed up without hiring for it, and that already have somewhere for the results to go. It earns its place when call volume is high enough that triage is real work, and when the follow-up connects to a CRM rather than a voicemail box nobody opens.
What to plan for before you commit. The technology is the easy part. Three things need a decision up front:
- Recording and consent law. Recording and AI processing of calls is regulated, and the rules differ by state and country, with some requiring every party to consent. Your greeting, your retention period and your privacy policy all have to reflect what you actually do. Get this wrong and the exposure is far larger than the software cost.
- Business messaging registration. Sending SMS as a business is not a switch you flip. US carriers require registration of the brand and the campaign, it takes time, and unregistered traffic gets filtered or blocked.
- The hand-off to a human. Decide when the agent stops and a person starts, and make that path obvious to the caller. An AI agent that will not let someone reach a human turns a good call into a complaint.
Visit ElevenLabs (affiliate link)
ElevenLabs is a trademark of its owner, used here under its affiliate program. We are an affiliate, not a reseller or certified partner. Features and pricing are theirs and change over time.
More to come
[PLACEHOLDER: add further tools only as they are genuinely adopted. Candidates to consider, each needing its own honest “where it fits and where it does not” note: the forms plugin, the email sending service, the scheduler, the analytics stack, the backup tool.]
Full detail on how this works: our affiliate disclosure.
Want the recommendation without the research?
Choosing infrastructure is a small part of what we do. If you would rather someone made these decisions in the context of your whole stack, that is an engagement, not a link.
