Govern every action an AI agent takes on your behalf.

The Agent Governance SDK resolves the authority behind every AI-agent action — an admissibility decision before execution, and a signed Evidence Pack™ after. Your agent proposes; the governance engine decides; the SDK returns a verdict you can act on and a record anyone can verify.

• Decision before execution. Every agent action is checked for admissibility — purpose, scope, and delegated authority — before it runs.
• Signed Evidence Pack™ after. Every governed action emits a replay-provable Evidence Pack™ — verifiable from published keys.
• An open wire contract, not a black box. The request/response shape is a published, versioned contract under Apache-2.0. TypeScript · Python · Go.
• Runtime-agnostic. The SDK speaks a published contract; the decision runtime is KYE Protocol™. Verify every record from public keys alone — no access to internals required.

• It is a client for the published governance contract: your agent proposes an action, the SDK returns an admissibility verdict and a verifiable Evidence Pack™.
• It is not the governance mechanism. This repository publishes the contract surface only (schemas, client types, examples). How decisions are computed and how evidence is signed are not part of this repository.

npm install @agent-sdk/agentgovernancesdk        # TypeScript / JavaScript
pip install agent-agentgovernancesdk             # Python

const verdict = await client.govern(action);
// verdict.admissibility → "admissible" | "refused"
// verdict.evidencePack  → signed, replay-verifiable from public keys
if (verdict.admissibility === "admissible") { /* proceed */ }

• Home: https://agentgovernancesdk.com
• Ecosystem: https://kyeprotocol.com
• License: Apache-2.0

Part of the KYE Protocol™ Agentic Governance™ ecosystem.