Software Engineer II - Cloud-Native Agent Harness (Mecatl)

About the role

Stacklok is led by CEO Craig McLuckie and CTO Joe Beda, two of the creators of Kubernetes. Kubernetes helped give enterprises more freedom over where and how they run software. Now we're bringing that same open approach to AI agents.

As AI agents take on more autonomous, long-running work, enterprises need infrastructure to run them securely and reliably at scale, without giving up control of their data, systems, or models. We're building open, Kubernetes-native infrastructure that brings together agent execution, secure access to tools and systems, and centralized controls for identity, permissions, and policy. This is an opportunity to help shape a foundational layer for enterprise AI while the architecture is still being defined.

We build in the open alongside the community. Our open source projects include ToolHive, an MCP platform that secures more than 20 million tool calls each month, and Mecatl, a cloud-native agent harness. Our commercial offering, Stacklok Enterprise, extends these capabilities with an AI gateway, centralized identity, policy enforcement, and observability.

Location

This is a remote role based in North America, with a strong preference for candidates located in the U.S. Eastern Time Zone.

The Opportunity

AI agents are moving into production, and the harness that runs them is becoming critical infrastructure. Enterprises need that harness to be secure, auditable, and portable across models, clients, and deployment environments. Mecatl is Stacklok's open source, cloud-native agent harness. It provides the agent loop, tools, permissions, hooks, delegation, and durable sessions for running agents as production workloads on infrastructure you operate.

As a Software Engineer II on the Mecatl team, you will build the platform that agents run on, and you will build it with agents. This is a hands-on role with real independence. You will deliver features and fixes across the engine, the Kubernetes-native runtime, the client surfaces (gRPC, HTTP/SSE, the TypeScript SDK, and the terminal client), and the permission and governance layers that make agent behavior safe to trust. You will go deep in one area while developing a system-wide understanding of how the harness fits together.

We are also an AI-first engineering team. We use agents throughout our own development lifecycle, and we are still working out what good looks like. Your own experiments, and what you learn from them, will feed directly into how the team works.

What Success Looks Like: First 3-6 Months

- Ramped quickly on the Mecatl codebase and its development process, and began contributing independently without needing close guidance.

- Shipped your first features and fixes within the first few months, then delivered core harness work consistently thereafter.

- Owned a product-minded improvement end to end, from understanding how people use Mecatl to shipping a change and learning from the result.

- Contributed a concrete improvement to how the team uses AI in development, such as a workflow, evaluation, guardrail, or measurement approach that the team adopted.