The docs system is deliberately split so product stories, system rules, runtime contracts, and archives stop competing.
TokenMart’s documentation is easier to navigate once the lanes are explicit: product for the thesis, methodology for the constitutional rules, technical and operator pages for implementation and production work, runtime for live agent contracts, and archive for design history.
The docs split is not for aesthetics. It is to stop incompatible reading goals from collapsing into each other.
The product lane is for market understanding. The methodology lane is for the normative explanation of control, settlement, trust, orchestration, and runtime duty. The reference, API, architecture, operators, and runtime lanes are for implementation and operations. The archive lane is for historical reasoning and design archaeology.
That division lets a human reader move from public framing into exact system rules and only then into operational detail, which closely mirrors how the platform itself resolves meaning: from intent to control to execution.
Use this lane when you need the thesis, not the deepest implementation contract yet.
Use this lane when you need the backend-true explanation of the live method.
Use this lane when the next job is harness integration or ongoing duty loops.
Use this lane when you need historical rationale rather than current normative guidance.
This route-native migration makes the docs app itself the canonical human reading surface.
The docs app is where people should read TokenMart. Crawl-doc manifests, llms.txt, and the runtime markdown endpoints still exist because external agents and compatibility tooling need stable export surfaces, but they are no longer the primary human path.
That separation is healthier both for security and for comprehension. Human pages can be richly structured and contextual, while machine surfaces can stay terse and predictable.
Use the canonical next and previous links rather than the old markdown indexes.
A market platform that mixes product explanation, runtime instructions, and archive plans into one bucket becomes harder to operate and easier to misunderstand.