Declare a discipline and its rules
Specify the domain, quality criteria, source hierarchy, policies and operational entities.
LowToHi documentation
Understand the system before you trust the system.
A practical map of the runtime, knowledge, authority, tools and evidence contracts that make LowToHi operational.
Operating model
Think of LowToHi as a compiler from a large capability universe to one finite governed execution.
Register models and capabilities
Connect local runtimes, APIs, knowledge sources, graph projections, Skills, MCP servers and tools.
Execute under identity and purpose
Resolve authority, compile context, route cognition and stop for approval where required.
Review claims, evidence and effects
Use receipts and evidence packages to reconstruct the execution.
Core concepts
Artificial Disciplinary Intelligence
A governed system that becomes exceptionally capable inside a declared professional discipline.
Persistent plane
Documents, records, graph relations, memory, Skills and outcomes available under provenance controls.
Execution snapshot
The finite, reproducible working set selected for one task.
Identity and effects
Permissions, purpose, policy, budgets, approvals and typed operations.
Bounded continuation
A workflow that continues only while its lease, budget and stop conditions remain valid.
Reconstructable evidence
Versions, inputs, decisions, costs, outputs and effects attached to one run.
Reference surfaces
Initialize, inspect and operate
Workspace setup, runtime checks, model registration, evidence export and diagnostics.
Submit and recover executions
Typed requests, stable identifiers, progress, cancellation, approvals and result retrieval.
Add professional capabilities
User-selected Skills and MCP connections enter through declared capability and authority contracts.
Admit context and effects
Versioned policy decides what can be activated, attempted and made effective.
Inspect claims and provenance
Claims, sources, graph relations and receipts remain addressable after execution.
Health, backup and recovery
Readiness, diagnostics, persistence, restart, backup, restore and rollback are first-class.
Architecture boundaries
- The model is replaceable infrastructure, not the authority root.
- Knowledge remains broad; task context remains finite.
- Browser clients never receive infrastructure credentials or hidden authority.
- Tools expose typed operations, not unrestricted channels.
- Improvement proposals and production admission remain separate.
- Unavailable capabilities are labelled, never simulated as live.
Low cost in. High trust out.
Build the intelligence your discipline actually needs.
Explore the product, inspect the evidence and see how LowToHi turns replaceable models into durable professional capability.