Give your agent authority you can prove.
KYE Authority Studio™ is the control surface where you take an agent — yours or off-the-shelf — and give it, bind it, and prove its authority. Register the agent as a first-class principal, attach an authority manifest, set an approval policy, wire a policy enforcement point into it, then watch every consequential action turn into a replayable Evidence Pack™. It governs agents; it does not build them — so a regulated buyer's security review drops from weeks to days.
Bring an agent, or scaffold a governed one.
KYE Authority Studio™ meets you where your agent already lives. Most teams arrive with an agent in production and need to govern it; some are starting fresh. Either way you leave with authority plus a verifiable receipt — never a black-box agent KYE™ built and now owns.
The differentiator: this is a studio that governs agents, not one that builds or runs them. The non-competition line with LangChain, CrewAI, the OpenAI Agents SDK and the Claude Agent SDK is deliberate — KYE™ governs the authority, you keep the agent.
Register → bind → policy → wire → prove → revoke.
KYE Authority Studio™ is a composition surface over KYE Protocol™ rails you already trust — the GovernedUI™ approval modules, the §52 authority bindings, and the §13 evidence and Authority Finality™ receipts. You move an agent through six honest steps, and each step is itself a governed, evidenced action.
- 1. Register the agent as a principal. The agent gets a canonical KYE™ identity and becomes discoverable in the directory — a first-class principal with the same governance weight as any human or system actor.
- 2. Bind its authority. Attach a §52 authority manifest, a Model Context Protocol allow-list, and a Purpose Permission™ scope — declaring exactly which actions, value bands, and data the agent may touch.
- 3. Set the approval policy. Choose an approval mode per action class — none, single approver, two-person, two-person-with-legal, delegated, or auto — reusing the GovernedUI™ Action Approval module, not a parallel model.
- 4. Wire the policy enforcement point. Generate a drop-in PEP adapter for your framework so the agent asks “is this action authorised?” the moment it acts — before money moves or a record changes.
- 5. Watch decision → evidence → finality. Every governed action emits the §0.3 evidence family and seals an Evidence Pack™ you can replay from public keys alone — Replay-Proof™, no access to your database required.
- 6. Revoke in one click. Pull an agent's authority the instant it drifts — the revocation is itself an evidenced, banking-grade act with dual-channel sign-off on irreversibles.
An AML agent closes an alert — and proves who let it.
Picture an AML triage agent your team already runs. When it closes a suspicious-activity alert or files a report, KYE Authority Studio™ decides whether that agent was authorised for that customer, that risk band, and that action — and seals the proof before the close lands.
- The action is checked at the boundary. The alert-close is sent to the decision point as a proposed action; Purpose Permission™ admissibility returns ALLOW only if the agent's bound authority covers it — otherwise it routes to a two-person review.
- The proof is sealed, not asserted. The verdict and its inputs are hash-sealed into an Evidence Pack™ in a write-once audit trail, re-derivable by your examiner via Replay-Proof™ against KYE Protocol™'s published keys.
- The value is faster exams. Because every agent action ships with mapped controls under EU AI Act™ Article 14, NIST AI RMF, and ISO/IEC 42001, exam-prep for a governed agent fleet drops from days of screenshots to minutes of replay.
A composition surface, not another silo.
KYE Authority Studio™ adds no parallel agent registry, no second approval model, and no build-your-agent runtime. It composes the canonical KYE Protocol™ rails — so what you govern in the Studio is the same authority your auditors, your directory, and your evidence trail already understand.
- Approvals reuse GovernedUI™. The Studio's approval policy is the existing Action Approval module — one canonical human-control surface, six approval modes, four channels.
- Authority reuses §52 bindings. Every agent's authority manifest and Model Context Protocol allow-list are the same §52 artefacts the platform governs everywhere else — no forked authority model.
- Evidence reuses §13 receipts. Decisions become Evidence Packs™ and Authority Finality™ receipts — the same replay-verifiable proof KYE Protocol™ emits for any governed action.
Honest scope: KYE Authority Studio™ governs whether an action was authorised and proves it. It is not an agent framework and never runs your agent — the output is authority plus a verifiable receipt, and nothing more.
Bring the agent. KYE™ makes its authority provable.
Start a proof of concept: register an agent, bind its authority, wire the enforcement point, and get back a sealed Evidence Pack™ for its first governed action — in days, not a quarter.