We're building the operational layer for MCP.
MCPG is a small team obsessed with one thing: making MCP traffic safe, observable, and operable at production scale. We chose Rust for the runtime, Apache-2.0 for the license, and an open RFC process for the design.
What we believe
- Protocol layer of agent infrastructure should be open. The gateway, plugins, control plane, operator, dashboard, and CLI are all Apache-2.0. A small set of enterprise modules ship under BSL-1.1 with automatic Apache-2.0 conversion three years from each release.
- Single binary beats microservices for ops layers. When everything else is on fire, the operational layer must work. One Rust binary is easier to reason about than ten Kubernetes deployments.
- Depth over breadth. AWS will out-distribute us. We win by being the deepest MCP protocol engineering shop in the market. We maintain a public conformance suite and we score against it.
- Dogfood-driven. mcpg.dev itself runs through MCPG. Our own control plane is an MCP workspace. If it breaks for us, it breaks for customers — that's a forcing function we don't want to lose.
Where we are
Public beta. The gateway runtime, 38 plugins, the control plane, and the Helm chart are running in production at design-partner orgs. The K8s operator and managed cloud are pre-GA. We're targeting 1.0.0 GA in 2026.
The team
A small group of engineers who've shipped policy engines, gRPC streaming protocols, plugin runtimes, and observability pipelines. We're hiring — see the contact page.
Working with us
Design partners get migration support, dedicated Slack, and direct input on the 1.0 roadmap. If you're running MCP at any non-trivial scale, we want to talk.