On Monday I published a post about lead magnets.
By the time I shut the laptop it had an opt-in box wired into it, an email sequence waiting behind that, the social posts drafted, and a record card filed so I’d remember what I’d done.
One session. One person. No team, no agency, no VA.
Five or six years ago that was a small team’s afternoon, and I want to be careful about why it isn’t any more… because it genuinely isn’t that I got smarter.
It’s the stack. And the stack is boring.
There’s no secret app in here. Most of these tools have been around for years, several are the obvious choice everyone already knows about, and most of them are free.
But here’s what surprised me when I actually added it up.
Running the website costs me about $41 a month. That’s hosting, the page builder, caching, images, the entire email system, links, SEO… everything that makes the thing exist on the internet.
The AI costs $120.
The thinking now costs three times more than the infrastructure. Five or six years ago that ratio was the other way round, and I’d have hired a person before I’d have spent that on software.
That’s the part every “best tools for solopreneurs” list leaves out. Those posts tell you what exists. They don’t tell you what one person actually runs, wired to what, for how much.
So here’s mine. Every tool, grouped by the job it does rather than the category it sits in, what each one costs, and the four I’d buy first if I were starting from nothing tomorrow.
Stack checked July 2026.
The short version
The useful unit isn’t the tool. It’s the job. Organise by category and you end up owning three things that all sort of do email. Organise by job and the whole thing stays legible.
Nine jobs run this business:
| The job | Main tool | What it’s for |
|---|---|---|
| Where the site lives | WordPress + Bricks | Publishing on something I own |
| Where the thinking happens | Claude | Drafting, structuring, deciding |
| Where words and pictures get made | Ulysses | Writing with nothing else in the window |
| Where email lives | FluentCRM | The list, the tags, the delivery funnels |
| Where I find out if it worked | Search Console | What I’m findable for, and what happened next |
| Where it gets out | Buffer | Social posts, drafted rather than scheduled |
| Where the work is remembered | Notion | Paying down the context tax |
| Where the day gets run | Todoist | The three tasks that matter today |
| Where I’m building | Next.js + Supabase | The product alongside the site |
Everything below is the detail behind that table: what each job actually needs, and what it costs.
1. Where the site lives
WordPress. I’ve been doing business solo since 2008 and I keep coming back to it. Not because it’s fashionable… because it’s mine. My URLs, my archive, my email list, all on hosting I control. Nobody gets to change the terms on me, raise the rent, or decide my format isn’t a priority any more.
Bricks. The visual builder that sits on top. Everything on this site is built with it: the homepage, the single-post template, the author card at the bottom of this page.
Novamira Pro. This is the one nobody has on their list, and it’s the reason my workflow looks the way it does. It gives Claude a direct, structured connection into the site… the pages, the Bricks elements, the design system, the settings.
A concrete example, because “AI builds my pages” means nothing on its own. I wanted an author card at the bottom of every post: photo, short bio, the lot. Instead of opening the builder, I said so. It read my single-post template, added the element, styled it against the existing design system, and told me it was done. Then I loaded a post and looked at it… which is the part that isn’t optional.
Page building stopped being an afternoon of dragging boxes and became a conversation. That’s not “AI wrote my copy.” That’s AI with hands on the actual site, and the difference between those two things is most of the last two years for me.
WP Rocket. Caching, minification, deferred JavaScript. The settings worth switching on are minify CSS, minify JS, defer JS and lazy loading. If you run a page builder, turn them on one at a time on a staging copy first, because aggressive JavaScript optimisation is precisely where builders break.
Imagify. Compresses images and serves WebP. Unexciting… and the cheapest page-speed win available.
2. Where the thinking happens
Claude, running two models for two different jobs. Fable 5 handles orchestration, the multi-step work where something has to coordinate several moving parts and keep track of where it is. Opus 5 is my daily driver for writing, thinking, and arguing with me about structure.
That split matters more than which specific models they are. One model to run the process, one model to do the work. The names will change. The shape won’t.
ChatGPT. My second opinion, and I use it differently from Claude on purpose. Research and fact-finding, early idea generation before I’ve decided what I think, design directions when I want options rather than an answer, and featured images for posts.
Running two assistants isn’t indulgence. They fail differently, and when both land on the same answer independently I trust it more than when one insists twice.
And the honest bit about how the writing works, because I’d rather say it than have you wonder. The drafts start as AI output. I rewrite the first-person sections in my own words before anything goes live, every time. No “humanizer” tools, no pretending it wrote itself and no pretending it didn’t. The drafts are raw material. The judgment is mine.
3. Where the words and pictures get made
Ulysses. Where I actually write. Plain markdown, no formatting bar, no comment threads, nothing blinking at me. Drafts live there until they’re ready to become a post, and then they move.
The reason this is a separate tool rather than typing straight into WordPress is simple: the writing app should have nothing in it except writing. The moment your draft lives inside the admin panel you start fiddling with the page instead of finishing the sentence.
Snagit. Screenshots and screen recordings. When I ran an experiment asking ChatGPT to recommend a business like mine, the proof was a screenshot of the actual answer… not my description of it. Same for any workflow I’m explaining: a recording of the thing running beats four paragraphs about the thing running.
It’s also what I’d reach for to send somebody a two-minute screen review of their site. That’s the kind of small, personal, unscalable thing no competitor with a bigger budget can fake.
Canva. Where the brand kit lives, for anything that needs laying out properly rather than generated.
4. Where email lives
The list is the most portable asset I own, so this part gets the most care. It’s also four tools rather than one, and the reason is worth understanding.
FluentCRM. Runs inside WordPress rather than as an outside service, which means my subscribers sit on my own hosting next to everything else. Tags, sequences, automations, campaigns. Somebody joins a list, a funnel fires, the thing I promised arrives without me touching it.
Contact Form 7. Catches the signup. Worth being precise here, because it’s the step people get stuck on: these two plugins don’t talk to each other on their own. FluentCRM’s native form integration is Fluent Forms, not Contact Form 7.
So I bridge them myself. About forty lines in a must-use plugin, hooked to wpcf7_mail_sent, calling FluentCRM’s API to create the contact and apply the right tag. The tag is what actually starts the delivery funnel.
You could do the same job with a connector plugin or an automation tool. I went with the snippet because it’s one file I can read in thirty seconds, and because the fewer moving parts sit between a signup and a delivery email, the fewer places it can silently break.
Mailgun. This is the piece almost every “best email tools” post leaves out, and it’s the one that quietly decides whether any of the above matters. FluentCRM stores and schedules your email. Sending it is a separate job.
You can push email straight out through ordinary web hosting, and plenty of people do. It just tends to give you weaker deliverability, thinner logging, and more authentication work to do yourself. A dedicated sending service handles the SPF and DKIM setup properly and tells you whether a message was genuinely accepted rather than simply “sent.”
WP Mail SMTP is what actually routes it. FluentCRM hands the email off, WP Mail SMTP sends it through the Mailgun API rather than through the web server, and Mailgun’s webhooks report bounces and complaints back so the list stays clean.
Mailgun’s free tier covers 100 emails a day, which is more than enough at my size. If you’re running email off your own WordPress install and haven’t set up a sending service yet, that’s a good hour to spend this week.
Pretty Links. Every download goes out through a /go/ link rather than a raw URL, so I can see which placement inside a post is doing the work… the inline box, the popup, or the line in the close. Guessing gets replaced with counting… which is the whole game.
Worth naming the pattern rather than just the tools, because the pattern is the reusable part: an inline opt-in box inside the post, plus one exit-or-scroll popup, both pointing at the same form. The popup fires at ten seconds or 35% scroll, shows once per visit, snoozes for a week if ignored, and disappears entirely once somebody has signed up.
I run two magnets through that pattern. The Simple Startup Canvas on posts about planning and positioning, the Free Startup Toolkit everywhere else. Same box, same mechanics. Different form, tag and funnel behind it. I clone it onto each new post rather than building a new one, which is the only reason it’s still consistent after a dozen posts.
One tip that costs ten minutes and beats everything else in this section: complete your own signup form, from a fresh browser, with a real address, all the way through to the delivery email landing in your inbox. Most people who build a funnel have never once run it end to end.
5. Where I find out if it worked
Three tools, because they answer three different questions.
Google Search Console. What Google thinks of me. Impressions, queries, average position, and specifically the queries where I appear at position 14 and nobody clicks… which is where my next post usually comes from. Free, slow, and it doesn’t flatter you.
Google Analytics, through the Site Kit plugin. What people did once they arrived. Site Kit is the part worth calling out: it pulls Search Console and Analytics into the WordPress dashboard, so the numbers are in the same place as the writing rather than in a tab I have to remember to open.
StatCounter. What one specific person did, step by step.
This is the one people ask me about, because on paper it looks like a duplicate of Analytics. It isn’t, and the difference is the reason I keep both.
Analytics is built for patterns. Totals, averages, segments, trends… it tells you what a thousand visitors did on average. StatCounter keeps the old-fashioned visitor log: I can open a single visit and follow the actual path. Landed on this post, scrolled, moved to that page, clicked this exit link, left from there.
Page view activity and exit link activity at the individual level is something I’ve never got out of Analytics in the same way. And for a business my size it’s more useful, because at low traffic an average of forty visitors tells you almost nothing, while one real path tells you a lot. You watch somebody arrive on a blog post, read two more, and leave through a specific link… and now you know something concrete about what to build next.
Worth being clear about what that is and isn’t, because “I can see the exact visitor” sounds more invasive than it is. These are anonymous sessions, not people… no names, no email addresses, nothing that tells me who you are. I’m reading a path through pages, not a person. If you run analytics of any kind and your readers are in the UK or EU, you still need a consent banner and a privacy policy that says what you collect, and both of those tools have retention and anonymisation settings worth actually opening.
The principle underneath all three: Search Console tells you what you’re findable for, Analytics tells you the pattern, and StatCounter tells you the story of one session. None of them substitutes for the others.
6. Where it gets out
Buffer. X and Facebook, with everything saved as a draft. Nothing posts until I’ve approved it.
Substack and Medium. WordPress goes live first, always. That’s the SEO home, and Substack has no canonical link support, so the original has to exist on my own domain before it exists anywhere else. Medium gets its import tool rather than a copy and paste, because the importer sets the canonical link back to my site automatically.
The rule I’d pass on: one home, many copies, and the home is the one you own.
7. Where the work is remembered
Notion. Every post gets a record card in one database: the URL, the WordPress post ID, the focus keyphrase, which lead magnet it feeds, which posts it links to, and what’s still outstanding on it. Featured image missing, social drafts unapproved, that sort of thing.
This looks like overhead… it’s the opposite.
When you work with AI agents, every session starts from zero. There’s a real price for that, and I call it the context tax: every restart where you re-explain your business costs me about twenty minutes. Across a week that’s the single largest hidden cost of working this way.
Notion is where I pay it down. Decisions written once, structures written once, learnings written once… boring, and it compounds. The agent gets better because the notes get better, not because the model does.
If you take one structural idea from this whole post, take that one. It’s free, and almost nobody does it.
8. Where the day gets run
The oldest part of my stack, and the part that hasn’t needed changing.
Todoist. I schedule my primary three tasks for the day. Not twelve… three. If the three are done, the day worked.
Google Calendar. Todoist is integrated with it, so those tasks appear on the calendar alongside everything else. I’ve tried a paper planner and I’ve tried Outlook, and I keep coming back to this pairing because what you put on your calendar is what actually gets done.
Three to four focused hours a day is what I’ve got. That constraint is the reason the rest of this stack looks the way it does… every tool above either saves a decision or removes a handoff.
9. Where I’m building
I’m building a product alongside the site, inside those same three to four hours.
Next.js, using the App Router, is what it’s built in. Supabase holds the data: the tables, the auth, the queries. Claude writes the code.
Playwright is the piece I’d argue hardest for. It screenshots and click-tests the interface in a sandbox before anything reaches me, which fixed a genuinely bad habit… I was shipping visual changes blind and judging them from a written description instead of looking at them. An agent telling you a page looks good is not the same as a page looking good.
None of that stack is unusual. What’s unusual is that I’m not a developer, and it doesn’t matter as much as it used to. I describe the behaviour I want, the code gets written, and the screenshots tell me whether it’s true.
Not saying more about the product yet… it doesn’t have a public name and I’d rather show it than trail it.
The rules that matter more than the tools
Swap every tool above for a competitor and this business would run about the same. Break one of these and it wouldn’t.
Nothing publishes without me. Social posts, email campaigns, live page edits. Agents draft, I approve. Approval is where my judgment enters the system, and automating it away would automate away the only thing that makes any of this mine.
Verify after every write. When an agent edits the live site, the job isn’t finished when the tool reports success. It’s finished when I’ve loaded the page and looked at it… every time. Tools report success on changes that quietly broke something more often than you’d expect.
Staging for anything structural. A separate staging site for the risky changes. Cheap insurance, and I use it.
Write the learnings down. See the context tax above.
What it costs
Here’s the whole thing, with list prices as of July 2026. Prices in USD unless noted, taxes excluded. Annual licences are shown as their monthly equivalent, so there’s a little rounding in the column.
The thing that surprised me when I laid it out: most of this stack is free, and the free tiers aren’t crippled ones. They’re free because a one-person business doesn’t hit the limits that trigger paid plans.
| Tool | Job | What it costs |
|---|---|---|
| KnownHost | Hosting | $25/mo |
| Bricks | Page building | from $79/yr (~$6.60/mo), $599 lifetime |
| Novamira | AI access to the site | Free (open source) |
| Novamira Pro | The Pro abilities | from EUR 49/yr (~$4.50/mo) |
| WP Rocket | Speed | from $59/yr (~$4.90/mo) |
| Imagify | Images | Free to 20MB/mo, then from $5.99/mo |
| FluentCRM | Free (I’m on the free version) | |
| Mailgun | Email delivery | Free (100 emails/day covers me) |
| WP Mail SMTP | Routing mail to Mailgun | Free |
| Contact Form 7 | Signups | Free |
| Pretty Links | Link tracking | Free; Pro from $249/yr |
| Yoast SEO | SEO | Free |
| Google Search Console | Search data | Free |
| Google Analytics + Site Kit | Behaviour data | Free |
| StatCounter | Visitor paths | Free tier |
| Claude | Thinking and building | Max, from $100/mo (Pro is $20) |
| ChatGPT | Research and images | Plus $20/mo |
| Ulysses | Writing | $39.99/yr (~$3.33/mo) |
| Snagit | Screenshots and recording | $39/yr (~$3.25/mo) |
| Canva | Design | Free tier |
| Buffer | Social | Free to 3 channels, then $5/channel/mo |
| Notion | Memory | Free tier |
| Todoist | The day | Pro, about $5/mo |
| Google Calendar | The day | Free |
| Next.js | The product app | Free, open source |
| Supabase | Product data | Free tier, then $25/mo |
| Playwright | Testing the product | Free, open source |
Roughly $170 a month, and $120 of it is AI.
Split it the other way and the shape gets clearer. Everything that makes this business exist on the internet… hosting, Bricks, WP Rocket, Imagify, FluentCRM, Mailgun, Contact Form 7, Pretty Links, Yoast, all three analytics tools, Buffer, Notion, Canva, Supabase… comes to about $41 a month. Several of those are free and none of them is expensive.
Then there’s Claude on the Max plan at $100, and ChatGPT at $20. That $120 is my two general-purpose AI subscriptions… the ones I think in. Novamira Pro sits in the infrastructure column instead, because it isn’t a model, it’s the connector that lets Claude reach the site.
I want to be honest about why I’m on Max rather than the $20 Pro plan, because I don’t want anyone reading this to think it’s the price of entry. It isn’t. I’m on Max because I’m building software with it, not because writing a blog post needs it. If you’re publishing content and running email, the $20 tier does that comfortably, and my bill would be about $90 a month instead.
But the ratio is the thing worth noticing, and it’s new. The tools that run a business have never been cheaper. The thinking has never been a line item at all. For most of the time I’ve been doing this, software was the expense and thinking was free because I did it myself. That’s inverted, and it happened fast.
One convention worth keeping if you publish your own: date your prices. Every one of the numbers above will be wrong within a year and you won’t be the one who notices. A single “checked July 2026” line turns a stale claim into an honest one.
If you’re starting from scratch, buy in this order
That table is twenty-six rows long and you don’t need twenty-six tools. You need four decisions, and the rest get added when a specific problem shows up and asks for them.
Decision one: where it lives. A domain and hosting. Somewhere you own.
Decision two: how you keep and send email. That’s two pieces, and the second is the one people skip.
Decision three: how you capture an address. A form, and something worth trading an address for.
Decision four: which AI you learn properly. One of them. The twenty-dollar tier is enough, and don’t copy my bill.
That’s a functioning business.
And here’s the part I’d want somebody to tell me.
The writing and the website are what earn somebody’s attention in the first place, and nothing works without them. But attention is borrowed. The form, the email tool and the sending service are the only pieces that turn borrowed attention into something you keep.
Everything else is scaffolding. Good scaffolding, necessary scaffolding, but scaffolding. When you’re deciding what to set up first, start with the three that convert a reader into a subscriber, and let the rest wait until they’re genuinely in your way.
The stack isn’t the business… it’s just what the business is standing on.
If you want the templates and AI workflows I use to run this one solo, they’re in the Free Startup Toolkit.
FAQ
What is a solopreneur tech stack?
The set of tools one person uses to run a business end to end: publishing, email, capture, measurement, distribution, and increasingly AI. The useful version is organised by the job each tool does rather than by category, because that’s what stops you owning three tools that half-solve the same problem.
What tools does a solopreneur actually need to start?
Somewhere to publish that you own, an email tool with a proper sending service behind it, one way to capture an address, and one AI assistant. Everything else gets added when a specific problem appears.
Why do you need Mailgun if FluentCRM already sends email?
Because storing and scheduling email is a different job from delivering it. You can send straight through ordinary web hosting, but it usually means weaker deliverability, thinner logging, and more authentication work to do yourself. A dedicated sending service handles SPF and DKIM properly and tells you whether messages were genuinely accepted.
Is WordPress still worth it in 2026?
For me, yes, and the reason is ownership rather than features. My archive, my URLs and my list sit somewhere nobody else sets the rules. Paired with a builder and an AI assistant that can work directly inside it, the old complaint that WordPress is slow to build in has mostly gone.
How much should a one-person tech stack cost?
Less than people expect on the infrastructure, and more than they expect on AI. A domain, hosting, an email tool and a sending service can genuinely run at well under fifty dollars a month, and several of those pieces are free at one-person volume. The AI subscription is now the largest single decision: twenty dollars a month covers writing, research and email work comfortably, and you only need more than that if you’re building software.
Which AI model should I use?
Less important than how you use it. What’s worked for me is splitting the job: one model to orchestrate multi-step work, one as a daily driver for writing and thinking, and a second assistant for research and second opinions. Pick one, learn it deeply, and write your context down so you’re not re-explaining your business every session.
Do I need to be technical to run this setup?
No, and that’s the real change. The bottleneck used to be knowing how to build. Now it’s knowing what you want clearly enough to describe it in a sentence.
