One engine. Two pieces that ship together.

KB Labs is an open-source runtime for the modern dev loop. Workflows turn releases, reviews, QA, and AI agents into code. The infrastructure layer keeps every external vendor behind a single contract. Both ship in the same install, on the same machine.

Programmable dev loop

Your releases, reviews, and AI agents as code.

We replace fragile shell scripts and CI YAML with one programmable system for releases, code reviews, QA gates, commit policies, and AI agent runs. Everything lives in code you can read, diff, and PR — the way the rest of your engineering already works.

  • One engine for release pipelines, review gates, QA regressions, and custom automation
  • AI agents as first-class steps — sandbox, audit, and observability built in
  • Workflows written in code, shipped as plugins, versioned in git
  • Designed to live alongside your existing CI, not replace it
How workflows work →
Infrastructure layer

Every external vendor behind one contract.

LLM, cache, databases, vector stores, object storage, event bus, observability — all reached through stable interfaces in the platform. Around 21 open-source adapters ship today, covering OpenAI, Postgres, SQLite, MongoDB, Redis, Qdrant, Docker and more. Swap any of them with a single config line; the code in your services stays untouched.

  • Around 21 open-source adapters today across the most common categories
  • Need Kafka, RabbitMQ, or NATS? Write your own adapter against the same contract — it loads identically to ours
  • Multi-tenant routing, audit, and observability built into the layer
  • Zero direct vendor dependencies in the core
See the contracts →
How they fit together

One engine, one runtime, one install.

Both pieces ship in the same package — same runtime, same plugin system, same on-prem deployment. The engine stays stable while the vendor layer underneath remains replaceable. That's how we make the next migration a sprint instead of a quarter.

Go deeper

How specific pieces of the engine work — and what people use them for.

Want to try it on your own workflow?

Install KB Labs on your own machine, point it at one process you actually own, and see how the engine fits.