Get in touch with Stacklok

Stacklok offers an open architecture for enterprises that need to run and govern always-on, highly advanced AI agents on their own infrastructure. Stacklok’s architecture includes the gateways you need to secure every agent action, including agent-to-tool calls (via our MCP gateway) and agent-to-model calls (via our LLM gateway). And Stacklok’s cloud-native agent harness enables enterprises to run and manage fleets of agents while protecting against lock-in to any one model prodivder (own the harness, rent the model).

Stacklok’s platform extends the capabilities of our open source MCP platform,ToolHive and cloud-native harness, Mecatl. Our CEO and CTO are the co-creators of Kubernetes and we’re a community-minded team. We believe in building in the open because we know that tech leaders value optionality and transparency.

Running MCP servers directly leaves significant gaps in security and usability. The MCP specification only addresses front-end authentication; it doesn’t cover back-end authentication, authorization, network isolation, or token exchange. Stacklok fills those gaps while also making it easy for admins to pre-approve and pre-configure servers, so end users get access to the context they need with a single click.