AXLEKEYSBETAAxle Keys is being built in public, one practical layer at a time. This page is the only place on the site that describes things that don't exist yet — everywhere else states what works today.
The modelling core, the fabrication outputs, and the agent endpoint are built and in use. What's still ahead is the knowledge that makes each System worth choosing, and the public spaces where finished models meet buyers.
The model loop works: tell your own agent what to build, or import an object; shape it, verify it, and take it out as drawings, cut lists, exports, or a configurator — in Studio or over MCP. The current work is the layer that makes the output specific: Systems that carry how a thing is genuinely made, proven against real projects with the first Keyholders rather than declared finished on a marketing page.
The platform is live and the modelling engine is real. Everything after this builds on it.
This is the working core of the product today. It is why the rest of the roadmap is worth building.
One model, many artifacts. A finished model produces the files a shop actually builds from.
People work in Studio, agents work over MCP, and both drive the same model.
A System is a build methodology captured as real context — parts, rules, constraints, materials. Open Design works today. A named System is only worth releasing when it knows enough to be genuinely better than starting from scratch, and that is the work in progress.
Access is deliberately narrow while the product is proven with real work.
Create was given a language and every other workspace was given buttons. A legible drawing, a build sequence a person can follow, an animation that explains itself, a parameter panel that fits the buyer — these are the outputs an agent is worst at improvising and a workshop is best at knowing. The language moves that knowledge into Axle, so the agent writes ten lines and the workshop does the craft.
The embeddable configurator and a model's own public page at axlekeys.com/models/… already ship, so the engine and the first public surface exist. What is missing is discovery — the places a buyer finds the model without ever opening Studio or being handed a link.
Every CAD tool will let you bring your own model, so the model is not the advantage. The advantage is how little intelligence a correct, buildable object costs. Other tools give an agent better CAD tools; Axle Keys aims to give it less CAD work. This phase waits on the language and the Library — it is measured, not guessed.
A few principles that guide what gets built and in what order.
One complete, working, well-documented System is worth more than five half-built ones. A System is only released when it knows enough to beat starting from scratch — which is why Open Design ships today and the named Systems are still listed as work in progress.
Every decision about what the platform can and cannot do is a product decision. Systems work because they have rules. A good configurator exposes the decisions that matter and protects the technical relationships underneath.
Every feature is evaluated against whether it makes the platform more callable — by people in Studio, by developers via API, and by agents via MCP. Callability is the long-term product. Agents are more useful when they work inside structure.
The roadmap will keep changing as the Systems become stronger and real users push models from idea to output. That is the point of building it in public.