Changelog

Expanded policy authoring, strengthened execution-time constraint handling, standardized throttling behavior, and improved reason codes for review, alerting, and SDK handling.

Improved Keel SDKs with easier governed execution setup, retry and timeout handling, stronger permit checks, and clearer examples across supported languages.

Added clearer platform status visibility, lightweight health checks, stronger execution management controls, and improved abuse resistance.

Aligned governed execution behavior across supported surfaces and strengthened validation, usage accounting, and audit consistency for permitted workloads.

Strengthened service credential validation and expanded audit evidence for privileged operations.

Updated API examples and reference links so published documentation more closely matches current platform behavior.

Improved usage ledger reconciliation and export verification. Customers with affected finance workflows received direct revalidation guidance.

Keel began with a simple operating principle: AI execution should be governed before work is allowed to run. From the start, Keel has focused on permit-based decisions, reviewable evidence, and controls that help teams prevent unsafe or unnecessary AI spend before execution.