1. AI powered apps inside claude

Three things inside Claude Projects that most people have no idea you can do. First one, honestly, blew my mind when I discovered it. So you can build a nap inside Claude where the app calls Claude's own API to do AI powered things. Basically, the app is calling itself. You can create a tool right inside the chat that has its own AI brain. Let's say I want a quick app that analyzes my Instagram comments and tells me what my audience is really asking for. And I built it as an artifact. So it calls Claude's API and it's an Audience Decoder. Paste in your Instagram comments, hit Analyze Comments, and it'll call Claude to group the comments by theme and suggest three video ideas based on what people are asking about. Make it look clean with a dark theme. This is for you. Let me first check the frontend design skill. So I created a file, read a file. Your Audience Decoder, paste in your IG comments, hit Analyze Comments, and it'll call Claude to group the comments by theme and suggest three video ideas tailored to your audience asking for. The 'sample' button pre-fills with 18 example comments. These samples are generated after the kind of comments you'd get on your code. Code content. The outputs include a title, opening hook, reasoning, and a dark theme. So it calls Claude's API and it's an Audience Decoder. Reading through comments... and it's a working AI powered tool that I built without writing a single line of backend code. The second one is something most people think is impossible. Persistent storage. Artifacts can save data that persists between sessions. Everyone thinks when you close a chat, everything disappears. But if you build an artifact with persistent storage, your data is still there next time you come back. So I could build a content tracker that logs what I post every day with the platform and the pillar. It should save between sessions so when I come back tomorrow my data is still there. No database needed. No backend. An artifact that remembers. And the third one is for anyone who builds across multiple repos. Claude sees your entire stack at once. So instead of Claude only seeing one repo at a time, you can connect more than one GitHub repo to a single Claude project. So instead of Claude only seeing one repo at a time, it can see your entire stack. I have my operating system repo, my skills repo, and whatever products I'm building, all loaded into one project. When I ask Claude to help me with something, it has the full picture. It knows my system, my skills, and the product code all at the same time. So these three things together turn projects from a chatbot into a full development environment. How to set all of this up, follow along.