Goal-first workflow
Users begin with what they are trying to accomplish, then CharterGate asks for context and an existing trusted source before preparing the real evaluation.
Windows private preview
About CharterGate
Start with the goal. Make the controlling rules visible before action.
CharterGate wraps a strict read-only governance evaluator in a guided product experience: begin in plain language, select an explicit trusted source, review the limits, and inspect the result and evidence before any separate execution.
Why it exists
Capability is not permission.
Modern agents and automated systems can do a great deal. That does not mean they should be allowed to decide their own authority. CharterGate is designed to place a deterministic governance checkpoint between a proposed consequential action and execution.
01Authority stays externalPermission comes from governed source-backed records, not model confidence.
02The exact proposal mattersActor, target, capability, scope, exclusions, conditions, timing, and provenance remain material.
03Execution stays separateCharterGate returns a result. Another authorized system may act later.
Product direction
Strict underneath. Friendly on top.
The product is being shaped so governance discipline does not require a governance engineer at every screen, while the source and authority boundary remain explicit.
Results, Sources & Evidence
People can understand the result quickly and still inspect the controlling source, binding, diagnostics, conditions, and audit record.
Non-actuating by design
Neither the interface nor Gate Keeper can turn a governance disposition into a business action.
Gate Keeper
Help without hidden authority.
Gate Keeper is the product's friendly guide for onboarding, help, result explanation, and fail-closed guidance. It can make the product easier to understand, but it has no authority over the governance result.
Presentation can be friendly. Authority resolution, source precedence, bindings, and dispositions stay strict.
AI integration model
Govern the proposed action, not the personality of the AI.
CharterGate can sit in front of many different agents or workflow systems because the evaluator focuses on the exact proposal and controlling authority rather than trusting a model's self-assessment.
Proposal
A human or AI-controlled application states the exact requested action.
Authority
Governed sources establish what permission actually exists.
Evaluation
CharterGate returns evidence and a governance disposition.
Execution
A separate authorized system decides whether to act.
Private preview
CharterGate is being productized now.
The current work is focused on the Windows private-preview application and the public product experience around the same read-only, non-actuating governance boundary.