AIToday
Large Language ModelsOpen-Source AIAI Business & IndustryLatent SpacePublished: Oct 8, 2026, 01:00 JST

Mecatl: Stacklok's cloud-native harness bet

Mecatl: Stacklok's cloud-native harness bet

3 Key Points

  1. What happened

    Craig McLuckie and Joe Beda's Stacklok released Mecatl, a cloud-native harness begun in June as an open source GitHub project, keeping the agent loop independent of the client, model provider, state store and execution environment.

  2. Why it matters

    Mecatl aims to move the locus of value from the desktop to the cloud, letting enterprises govern agent sessions the way they manage email rather than leaving code and context on individual machines.

  3. What to watch

    Beda says the 'lift and shift' approach creates problems around agent lifecycle and quiescing, so the test is whether separating the loop from sensitive operations scales across multiple clusters and clouds.

WHO IT HITSEnterprise platform and infrastructure teams in regulated industries such as banking, semiconductors and telecoms, who need to deploy agentic coding tools centrally rather than manage them per developer laptop.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

Stacklok's pivot from software supply chain security to Kubernetes-based agentic solutions is led by two people who helped create Kubernetes at Google: Craig McLuckie and Joe Beda. The company raised a $17.5 million Series A back in 2023 from Accel, Madrona and Bain Capital. Its earlier open source product, ToolHive, started as a way to manage MCP servers running in Docker containers on the desktop, then expanded into a Kubernetes-based gateway, registry and operator helper.

Mecatl's design follows the same reasoning behind Kubernetes control loops, which keep applications running, recover from failures and scale when needed. The pair argue that container orchestration in the early 2010s faced similar challenges around reliability and isolation. McLuckie frames the goal as running in a cloud decoupled from the specifics of the cloud provider, echoing what the team did in the Kubernetes era. Beda claims many agent tools were designed for individuals or small startup teams, and McLuckie says most Stacklok customers are banks, semiconductor companies and telcos in relatively regulated industries.

The stakes appear to hinge on whether enterprises will pay for a central control plane rather than accept the harnesses that arrive bundled with frontier lab tools. ToolHive and Mecatl are intended to be independently useful, so the open source projects are easy for a small team to run; Beda says scaling them out to multiple clusters and multiple clouds surfaces new problems. It remains an open question how quickly regulated enterprises move agent workloads onto Kubernetes, and whether Stacklok's commercial spine becomes the standard for doing so.

FAQ
What makes Mecatl different from desktop-first harnesses?
Mecatl keeps the agent loop independent of the client, model provider, state store and execution environment. Beda says desktop-first harnesses often begin the loop, local execution and session state inside the same process.
How does Stacklok make money if Mecatl is open source?
McLuckie says the commercial product is the 'enterprise spine, the control plane' that ties the open source parts together. It offers consistent identity, authorization, policy and auditing across them.
Is Stacklok's AI Gateway open source?
No. The AI Gateway is the one main product Stacklok hasn't open sourced yet, though Beda says it's 'on our roadmap and we're going to get there soon.' It focuses on access control, budgets, reporting and provider routing.

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleGoogle launches Playground for text-to-game creation