Over the past year, I have been making Home Inspection Report Software called SwiftReporter.
SwiftReporter Android
SwiftReporter iOS
SwiftReporter is an AI-assisted report-writer for Home Inspectors. It helps them create templates, comments, and observations for their reports quickly, so they can focus on the building inspection, and rather than spend time writing the report,s they can free up that time to use as they like.
The common reality in the industry is that when you start a new report software, you should expect to spend 20-40 hours creating your first template for your home inspection report. SwiftReporter has Agentic assistants and tools to help quickly create the template and overcome this major hurdle. I have used the agent to build a template for a 45-page report in only 15 minutes using my software. That’s a lot of time saved!
It’s free to use, although if you want to share more than a few reports, then payment is required. There is no CC required for signup, so if you want to play around with it, there are no obstacles there. I’d welcome any feedback!
The Template Agent in Action:
How I used Cursor
I started using Cursor when development began in earnest in about May, and things have really accelerated since then. I am just one developer, working about one day a week, and yet I’m able to push a new feature release each week across the website, web app, iOS and Android apps. I am a Senior Dev, and I’m not vibe-coding, but the way I use Cursor puts me very close. I find that I often need to guide the agent to structure the code in ways that I want, but lately I find I don’t need to review as closely as I did in the past, and I let the Agent loose on tasks that are not risky from a safety perspective.
Here are some ways that I use Cursor while developing:
Website - I use Webflow and the Webflow MCP server. This allows me to manage changelogs, docs, and even blog posts (AI doesn’t write the blog, but it does review/upload). I have /commands to write a local changelog for each of my releases, the Agent gathers the git diff, and writes a summary, then posts this to my changelog in Webflow (take note, Cursor team 
). The same goes for usage documentation, this helps me keep it up to date when otherwise this would be the first thing I would fall behind on.
Web App - The main web-app is a Next.ts/React app, written entirely with Cursor. I hadn’t actually used Next.js before, and I have many years of React/Typescript experience. I really depended on the Agent to help me through many of the little configuration nuances of setting this framework up. It ended up being a pretty seamless experience.
iOS/Android/Backend - An app is essential for this software, because the Inspector’s often work in places with no connection to the web, and they need a stable platform that can work offline and sync with the cloud when the connection is re-established. I decided to use Firebase and React Native. I chose Firebase because of Firestore’s native support for offline mode and serverless architecture, but I have never used Firebase before, and the learning curve was steep (AWS and GCloud, yes). I leaned on the Cursor Agent for this as well, and it worked brilliantly. Especially with the later model Gemini AI. I can ask the Agent to utilise the cli for anything related to Firebase, and it will reliably perform checks that I’d have to comb the docs for if I was on my own. I use the Firebase MCP and the Cursor Agent to do a weekly analysis of my users’ activity. It identifies which users I should prioritize a follow up email with, where they might have gotten stuck in the process, and how seriously they are engaging with my app.
As for iOS and Android, React Native felt the most accessible because I know React, but I have never made an app before, so AI really guided me through configuration challenges here, and helped me understand issues that I ran into.
Workflow - When using Cursor, I take advantage of rules, commands, and create inline documentation (written by the agent). Whenever I encounter something that the Agent doesn’t do to my liking, I create a new rule to specify how I want it done. This doesn’t happen as often as you might think, and each repo has about 10 rules that explain things like styles, shared components, folder structure, etc.
I use commands for anything I do often. My favourite is called /move , and it contains instructions to refactor the selected code into a new file. It also contains an explanation of the folder structure and naming conventions, so the new files are created correctly and in the right place without me needing to explain.
Whenever I make a new commit, I have the Agent write documentation in a folder in the repo called docs/. This is minimal, and the AI is instructed to write the docs “for use by AI Agents”. I find that having this inline documentation provides the AI with business logic and reasoning for the ‘Why’, rather than the ‘How’ that it can get from reading the code itself. This is then indexed by Cursor and is returned whenever the Agent performs searches in the codebase using vectors.
I have some great hooks that I use as well that let me let the Agent loose and prevent it from deleting files without my approval, plays a tone when it’s waiting for me to approve something (the Cursor tone is only when the agent finishes), runs my formatter and runs the linter. I’ll share this in another post if people are interested. I think some of it might have been made obsolete by the Sandbox mode, but I haven’t tested that yet.
I’ve also used Cursor Agents to help me build search tools to collect the email addresses of almost every Home Inspector in the US (very helpful!). I’m looking forward to building some more back-of-house tools to help me more with responding to user feedback and triage, but I’ve been so focused on building that I’m just now getting the capacity for that.
Looking forward to what I’ll be able to build this next year!
Links: 
Star the repo if you find it useful! Feedback and contributions welcome.
. I realize i drop a lot of bugs in the bug forum, but holy cow this newfangled agent-based coding is the coolest thing since i made the big step from flipping front-panel switches on a PDP-10 to emacs.
Cortex Memory - Plug’n’play persistent memory for AI agents
Production-ready v0.28
124 test files with 18,460+ assertions
Full security scanning (CodeQL, Semgrep, Trivy, Gitleaks, Bandit, OpenSSF Scorecard)
Complete documentation site at docs.cortexmemory.dev
TypeScript + Python SDKs
Powered by Convex for real-time reactivity
Hive Mode - Multiple AI tools share one memory (your Cursor agent + Claude + custom tools = zero duplication)
Belief Revision - Automatically updates facts when users change their mind
Graph Integration - Optional Neo4j/Memgraph support
GDPR-ready - One-click cascade deletion, compliance templates
Interactive demo - Live visualization of memory orchestration with Vercel AI
docs.cortexmemory.dev
npm: 
