How Kong Built an Agentic Engineering Platform

Kong developed KSAI, or Kong Skills for Agentic Intelligence, as a GitHub Actions-based platform for asynchronous AI-assisted code review and implementation, replacing fragmented individual tooling with reusable, governed workflows. KSAI routes model traffic through Kong AI Gateway to provide vendor-neutral model selection, per-repository usage attribution, budget enforcement, observability, and fallback capabilities, while sandboxing agents with read-only filesystems, restricted permissions, isolated namespaces, removed runner credentials, and separate authorization and approval controls. After finding that independently operating review subagents duplicated repository exploration and token costs, Kong introduced a static triage phase that selects whether to review a change and which model and effort level to use, reducing median review costs by about 37 percent. Internal benchmarks evaluate models through blocking-defect recall, precision, F1 score, and cost-weighted scoring; although Claude Opus performed strongly, GLM 5.3 and GLM 5.3 Flash were selected as default reviewers because they offered similar quality at substantially lower cost. By September 17, KSAI had been adopted by 22 teams across 135 repositories, reviewed 7.8 million lines of code, produced more than 4,000 findings, and supported self-service onboarding and automated workflow upgrades, while also being used to review and help develop its own codebase.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.