MachineFabric is a local-first semantic engine for macOS. It transforms the files you have into the insights you need by weaving together a fabric of modular intelligence.
Operating systems see files as mere bytes on disk. MachineFabric sees them as latent potential. Right-click a PDF and it doesn't just launch an app, it reveals what that file is capable of becoming: a page by page summary; a set of embeddings, the answer to a question, or an audiobook.
You choose the destination. The Planner resolves the route, automatically chaining the necessary Cartridges to carry the file from what it is to what you need.
Right-click to routed output
Temporary placeholder until we replace it with a MachineFabric capture showing file intake, route selection, and transformed output.
Intelligence shouldn't require a data lease. MachineFabric runs entirely on your hardware. A Relay Architecture isolates processing into sandboxed Cartridges—your files never leave your machine.
Once your models are local, you are truly offline. No cloud APIs. No subscription gates. No data harvesting. Sovereign Intelligence: the power of modern AI with the security of a local binary.
On-device processing surface
Reserved for a MachineFabric loop that shows local execution, private artifacts, and zero cloud dependency.
Stop building wrappers. Start building Cartridges. A Cartridge is a self-sufficient, POSIX-compliant binary that communicates via the Bifaci Protocol.
Every Cartridge is self-describing. Plug it in and it advertises its Tagged URNs—the system's vocabulary for what it accepts and what it produces. Whether you write in Rust, Swift, or Go, your tool becomes a first-class citizen in a global capability mesh.
Capability authoring preview
Reserved for a future workflow showing URN declaration, cartridge validation, and first-class registration in the fabric.
Behind the curtain, we aren't guessing. MachineFabric uses Category and Order Theory to route data. Through the CapDAG protocol, the system performs Semantic Dispatch—mathematically ensuring that a provider's capabilities perfectly refine your request.
If Machine Notation can describe a path from a Video Stream to a Summary, we guarantee the types will align and the data will flow. Transformations are deterministic and verifiable.
Typed route resolution
Reserved for a future query-to-route visualization with deterministic dispatch, matched capabilities, and verifiable execution.
We built MachineFabric for the way data actually moves. Streaming-native, with multiplexed Bifaci channels processing frames in real-time. No loading everything into RAM first.
Process live streams, massive PDF libraries, or high-fidelity audio. The Orchestrator handles memory pressure and execution topology—our job is to handle the plumbing so you can focus on the logic.
Streaming-native execution
Reserved for a future capture showing cartridge activity, flowing streams, and sustained throughput under load.
MachineFabric isn't just an app. It's a new layer of the macOS experience. An open protocol. A modular ecosystem. A private sanctuary for your data.