Kenshiki

Start governed synthesis without private infrastructure

Workshop

Shared Kadai or BYOK. Governed outputs.

Workshop is built for the moment a fluent answer sounds right and you hesitate. It runs the full Kenshiki governance pipeline on shared infrastructure. Generation can happen through Kenshiki-hosted Kadai models or a connected model API like GPT, Claude, or OpenRouter. The model still generates the language. Kenshiki controls what it sees, evaluates what comes back, and decides what reaches a user. In the three-plane architecture, Compiler and Kura handle build, Kadai orchestrates, Ledger and Gate enforce control — so cross-plane policy propagation has no seams. SIRE provides portable agent identity so evidence scope follows each caller across runtimes.

Without this: the model answers directly, and the burden of deciding whether it's usable falls on whoever reads it. You discover problems after someone relies on them.

Today

Your team calls a model directly. It returns fluent text. Someone reads it, decides it sounds right, and passes it along. When that output is questioned — in a review, an audit, a legal proceeding — no one can show what it was based on.

With Workshop

The same model can still generate the language — shared Kadai or a connected API — but now inside a governed loop. The prompt is compiled, evidence is retrieved, claims are checked, and the response gets an explicit state before it reaches anyone.

How Workshop works

A user question enters the Kenshiki pipeline before the generation layer sees it. The prompt is compiled, evidence is retrieved, and bounded context is sent through Workshop's shared generation lane: Kenshiki-hosted Kadai or a connected model API. The response comes back as a proposal, is decomposed into claims, checked against evidence, and assigned a state before it reaches anyone.

Kenshiki control plane · Signed envelope · Chain of custody
Your data · Outside Kenshiki

Output states

AUTHORIZED
PARTIAL
REQUIRES_SPEC
NARRATIVE_ONLY
BLOCKED
AUTHORIZED Claims sufficiently supported by evidence
PARTIAL Evidence exists but coverage is incomplete
REQUIRES_SPEC Question is underspecified
NARRATIVE_ONLY Descriptive but not decision-grade
BLOCKED Policy or evidence conditions not met

What Workshop is

The full Kenshiki bounded-synthesis pipeline on shared infrastructure. Compiler, retrieval, Claim Ledger, and output-state assignment all run inside Kenshiki. Generation can happen through Kenshiki-hosted Kadai or an approved model API, both receiving only the constrained context Kenshiki provides.

  • Use shared Kadai or a BYOK/public endpoint as the generation layer
  • The full governance pipeline runs in front of and behind that model
  • Answers are evaluated before they reach a user

The Kenshiki contract

Two APIs. One contract.

Put evidence into Kura. Ask Kadai for answers bounded by it. In Workshop, the generation layer can be shared Kadai or a connected model API — but the contract is the same. SIRE scopes what evidence is admissible. The generation model acts as a renderer inside the governance pipeline, not as the authority.

  • Kura defines the evidence boundary
  • SIRE scopes evidence admissibility per query identity
  • Kadai returns answers bounded by that evidence
  • The generation model renders — it does not decide

Who this is for

The Team Shipping AI Into Real Work

already using public model APIs, or wanting a shared Kadai starting point, but unable to justify what those systems produce under review, audit, or challenge.

The Decision-Maker

receives an answer only after Kenshiki has checked what supports it, what is missing, and whether it is allowed to leave the system.