About

We build infrastructure
the way we wish
someone else had built it.

Gateway-LLM is built by Mitshaw Technologies, a small engineering team that spent too long stitching together LLM providers in production and decided to write the gateway we wanted to use.

The first version of Gateway-LLM solved one problem: stop paying flagship prices for prompts that don’t need a flagship. The next versions added everything you needed to run that in front of real customer traffic — virtual API keys, budgets, audit, observability, multi-provider failover, all in one binary.

The product is open-source because the people picking the LLM gateway in 2026 are the same people who picked Postgres, Kubernetes, and Caddy in 2020. They want to read the code. They want to run it on their VPC. They don’t want to learn a new SDK. We respect that, and the product reflects it.

What we believe

  • Open-source by default
    The whole product is Apache 2.0. The hosted version uses the same binary. There's no community edition / paid edition split.
  • Engineering-first
    Our docs assume you can read a config file. Our SDKs assume you can read a type. Our pricing assumes you can do the math.
  • Price on outcomes, not on metering
    Per-token markups feel like vendor lock-in. We charge a percentage of the savings smart routing creates — when it doesn't save you money, you don't pay us.
  • Trust the user's stack
    We make the gateway disappear into your existing tooling. OpenAI SDKs work as-is. Datadog, Prometheus, Langfuse, your warehouse — same shape data they've always had.

Want to talk?

Email the founders directly. We answer fast — usually same day.

mitshawtechnologies@gmail.comRead the Quickstart