In 2026, one developer with AI tools can deliver what used to require a 5-person agency.
I'm not saying this as hype. I'm saying it because I've been living it. Over the past few months I've built 25 n8n automation templates, published 2 MCP servers, packaged 17 Claude Code skills, and started landing Upwork clients — all as a solo developer.
This is my honest breakdown of how the business works, what the stack looks like, and what the numbers actually are right now.
The Core Insight
Traditional automation agencies need:
- A project manager to scope work
- A developer to build the automation
- A QA person to test it
- A copywriter to document it
- A salesperson to find clients
With Claude Code + n8n + MCP servers, one person covers all five roles. Not perfectly. But well enough to deliver real value to real clients.
That's the wedge. And I'm using it.
The Stack
Claude Code — My Development Partner
Claude Code is Anthropic's official CLI. It runs in your terminal, reads your codebase, and helps you build at a pace that genuinely feels like having a senior developer beside you.
I use it for:
- Scaffolding new n8n workflow logic before building in the UI
- Writing TypeScript MCP servers from scratch
- Generating Shopify API integrations
- Debugging webhook failures
- Writing client-facing documentation
The key shift is treating Claude Code as a collaborator, not a code generator. You're still the architect. It handles the boilerplate, the syntax, the "how do I do this in TypeScript again" questions.
n8n — The Automation Engine
n8n is my core delivery tool. It's open-source, self-hostable, and has the best balance of power vs. usability I've found. Clients can actually look at workflows and understand them, which matters for handoff.
My workflow with n8n:
- Client describes a business problem
- I map it to an automation pattern I've already built
- I adapt one of my 25 templates to their specific setup
- Deploy, test, document, hand off
Having a library of battle-tested templates means I can deliver a project in hours instead of days. That's my competitive advantage.
MCP Servers — The Integration Layer
Model Context Protocol is how AI agents connect to real tools. I've built two production MCP servers:
- Shopify MCP Server: 16 tools for managing products, orders, inventory, customers, and analytics — all accessible directly from Claude
- Google Sheets MCP Server: 14 tools for reading, writing, formatting, and querying spreadsheets — filling a real gap in the market
MCP servers turn Claude into a genuine business tool. Instead of copy-pasting data between systems, clients can ask their AI agent to "pull last week's orders and update the inventory sheet" and it just works.
What I Actually Sell
1. Custom Automation Projects (Upwork)
This is the fastest path to cash. Clients post jobs, I apply, I deliver.
My positioning: "AI-powered automation using n8n and Claude Code." I'm not just an n8n developer — I bring AI into the workflow itself.
Typical projects:
- Order processing automation ($300-800)
- AI-powered email responders ($400-1200)
- Lead scoring pipelines ($500-1500)
- Voice agent integrations ($800-2000)
- MCP server development ($1000-3000)
Rates: $40-80/hr or fixed price. Still building the profile, but the pipeline is real.
2. n8n Template Pack (Digital Product)
I packaged 25 production-ready n8n templates into a bundle covering e-commerce, content, operations, and advanced AI use cases.
Each template is:
- Fully documented
- Ready to import
- Covers a real business use case
The bundle is live on Gumroad. Having it there also serves as a portfolio — clients can see the scope of what I've built before hiring me.
AI Business Automation Pack on Gumroad
3. MCP Server Marketplace
Both MCP servers are published and listed on MCP directories. The goal is passive income from developers who need these integrations without building from scratch.
The Google Sheets MCP fills a genuine gap — there's no dominant solution in that space yet.
4. Claude Code Skills
17 packaged Claude Code skills covering git workflows, code review, API scaffolding, project health checks, freelance proposal writing, content pipelines, and more.
Skills are a new distribution format — portable, composable, shareable. Think of them as specialized AI workflows you can install and invoke.
Revenue Channels (The Four-Channel Model)
I'm not betting everything on one channel. Here's the split:
| Channel | Type | Timeline |
|---|---|---|
| Upwork freelancing | Active income | Immediate |
| Digital products (Gumroad) | Passive income | Medium |
| MCP marketplace | Passive income | Medium |
| Content marketing | Lead gen | Long term |
The theory: freelancing pays the bills now while products build passive income over time. Content drives inbound for both.
Real Numbers (Early Stage, Honest)
I'll be direct: this is early. I'm not writing this from a beach with a $20K MRR dashboard.
What exists right now:
- 25 n8n templates built and packaged
- 2 MCP servers compiled and listed
- 17 Claude Code skills ready
- Upwork profile live, proposals going out
- Gumroad products created (awaiting payment setup to publish)
- 6 dev.to articles published (including this one)
- Multiple freelance platforms set up (Upwork, Fiverr, PeoplePerHour, Freelancer.com)
What I'm targeting:
- $2K-5K/mo in 90 days from freelancing
- $500-1K/mo in 6 months from digital products
- $200-500/mo in 6 months from MCP marketplace
The blueprint is sound. Execution is the variable.
Lessons Learned So Far
1. Marketing is the actual job
I spent months building things. Products that no one knew existed. The hard truth: a mediocre product with great distribution beats a great product with no distribution every time.
Now I spend more time on:
- Writing articles (like this one)
- Submitting to directories
- Posting proposals
- Cold outreach to potential clients
Building is the fun part. Marketing is the job.
2. Start with services, then productize
Every template in my bundle started as something I built for a real use case. Every MCP server solves a problem I actually hit. Services first forces you to build things that work in the real world, not things that are interesting in theory.
Once you have something proven, package it. Then you're selling the same work multiple times.
3. Templates accelerate everything
Having a library of working templates means:
- Faster project delivery (hours not days)
- Lower risk for clients (they can see what they're getting)
- Easier proposals (I know exactly what it takes)
- Passive income channel (sell the templates directly)
If you're doing automation work and not building a template library, you're leaving money on the table.
4. The niche matters more than the tool
"I do n8n automation" is a commodity. "I build AI-powered Shopify automation systems for e-commerce operators" is a positioning.
The more specific you are, the easier it is to find clients, write proposals, and charge a premium. I'm still refining mine, but the direction is e-commerce + AI + Claude Code.
5. MCP is undermonetized right now
The MCP ecosystem is exploding — 97M+ monthly SDK downloads, 10K+ active servers, 8 new directories launched in the last 6 months. But less than 5% of MCP servers are monetized.
That gap is an opportunity. Building MCP servers now, while the ecosystem is young, is the equivalent of building Shopify apps in 2012. Most developers are ignoring it because the monetization infrastructure is still forming.
The Tools List
For anyone who wants to replicate this:
Development
- Claude Code (Anthropic) — AI pair programming
- n8n (self-hosted or cloud) — workflow automation
- TypeScript — MCP server development
- Node.js/Express — API backends
Business Infrastructure
- Upwork — freelance clients
- Gumroad — digital product sales
- Polar.sh — developer-focused payments (4% vs Gumroad's 10%)
- GitHub — open source + portfolio
Distribution
- Dev.to — technical content
- MCP directories (mcp.so, glama.ai, PulseMCP) — MCP server discovery
- n8n marketplaces (n8nmarket.com, ManageN8N) — template distribution
- Fiverr, PeoplePerHour, Freelancer.com — freelance reach
AI Services
- DeepSeek API — cost-effective OpenAI-compatible API
- OpenAI API — when clients need it specifically
What's Next
The immediate focus is converting the infrastructure into revenue:
- Get first Upwork client (proposals going out daily)
- Publish Gumroad bundle (payment method setup pending)
- Get MCP servers listed on top 5 directories
- Publish this article and the five others in the backlog
- Submit to Shopify App Store (0% commission to $1M — hard to ignore)
If you're building something similar — a solo dev agency, a product business, anything in the automation space — I'd genuinely like to compare notes.
Leave a comment or find me on Upwork.
Resources
Building in public. Early stage. Real numbers when there are real numbers to share.
📦 Get My Templates
Want these templates ready to use? Check out my complete collection of
25 production-ready n8n automation templates — ready to deploy immediately.
👉 Get the n8n Templates Bundle ($10)
Perfect for automation engineers, consultants, and anyone looking to accelerate their n8n projects with battle-tested workflows.
Top comments (0)