I’ve been part of the forum for several years but not actively posting.
I recently started using Claude Code and, upon discovering its capabilities, I immediately thought about the possibility of connecting it to Blocs through MCP (Model Context Protocol). The idea would be to allow Claude to read and edit Blocs project files directly.
It seemed like a natural combination: bringing together a visual environment I know relatively well with the potential that Claude Code offers. For a designer who also needs to produce web design, this would be an almost unbeatable tool.
However, I quickly realized that Blocs uses a proprietary format (.blocs) that is not externally accessible, making the system closed to editing by external tools. Although Blocs exports standard HTML based on Bootstrap, any changes made to the exported code are lost on the next export, there is no bidirectional workflow.
I ended up building an alternative solution, but the question remains. Are there plans for Blocs to provide an MCP server or another form of interoperability with Claude Code that would allow access to the project format?
I searched the forum and this topic appears not to have been discussed yet.
That sounds futuristic, but that’s probably the direction things are heading in anyway. It remains to be seen where Blocs is heading in the future, but that would technically be quite complex if @Norm and his team would go that route.
For Blocs to stay true to its “no-code” roots, an integration like MCP (Model Context Protocol) would have to be incredibly seamless. If it becomes too technical, Blocs loses its greatest advantage: its intuitive, visual workflow. To provide real value without overwhelming the user, @Norm would probably have some really difficult and tricky problems to solve, I think (but I’m no AI expert).
It is more likely that we would first see small AI assistants within Blocs (e.g., for text generation or color schemes). But you’ll have to ask @Norm about that.
But let’s wait and see what the future brings (I hope lots of new features), and the next version (6.4) is already in the starting blocks.
I don’t know how much effort is required to integrate Claude Code.
But with Claude Code, as with all services, getting started is free, but if you want to do more, it costs money. I couldn’t find the prices for Claude directly, but it may well be that it costs more than the BlocsApp itself. The billions invested in AI have to be recouped somehow
It’s probably easier to integrate some Apple AI services…
Great points raised in this thread. I’d like to add a few thoughts that might help shape a realistic path forward.
On the cost concern: A practical approach would be to let users bring their own API key (Claude, OpenAI, or others). This way Blocs doesn’t absorb the cost, and users who want AI features pay for what they use. Many tools like Cursor and Raycast already work this way successfully.
On complexity: Full MCP integration doesn’t have to happen overnight. A phased approach could work well:
Phase 1 – Lightweight AI assistant inside Blocs for text, color suggestions, basic CSS help
Phase 2 – A simple API or plugin layer giving limited access to project structure
Phase 3 – Full MCP server if community demand justifies the investment
On Apple Intelligence: It’s a logical starting point since Blocs is Mac-only, but Apple’s AI capabilities are still quite limited compared to models like Claude. It could be a low-friction entry point, not the final destination.
One idea that could unlock a lot without major effort: publishing partial documentation of the .blocs format. Blocs wouldn’t need to open-source anything, just enough for the community to build integrations independently, reducing the burden on @Norm and the team entirely.
Curious to hear what others think, and of course what direction @Norm is considering.
I’m actually an avid user of Claude myself, especially useful for debugging issues (analyzing suspicious or potential malicious code string) within PHP projects in my day-job (DevOps engineer at a hosting company in Belgium) and I’m on the lowest paid tier for Claude, which suffices perfectly for most day. I rarely hit the limit on that tier. The cost is about 20 euro each month, so I still deem this to be justifiable.
Besides that, all great points made in this thread. Perhaps Claude AI MCP integration could be a feature in the next major release.
Good idea. I think that would be great. Not as open source, of course, just as information. Especially for larger projects, this would reassure some people.
That’s easy to say, but every developer knows that good documentation takes an enormous amount of time and every little detail has to be kept up to date…
Apple uses Gemini for it’s AI based foundation for Siri. At least for integratying AI into Apple’s Swift programming language, you can use ChatGBT or Claude. For Claude it needs to be a paid version of their service.
We have lots of plans for AI and I think having a way to let various AI models have more control of Blocs documents and data is a logical step forward.
But some things will need to be developed further in order to explore the full possibilities, exciting times ahead
I’m also now using Claude Code — not really for coding, but for a lot of other things — and this would be an absolute game changer. I find Blocs quite difficult as a non-developer, and coming from RapidWeaver, integration with Claude would be an absolute ripper and, honestly, the future.
