GitHub, Solo.io take on MCP; LoftLabs secures K8s nodes

Welcome to Runtime! Today on Product Saturday: more companies line up behind MCP, which could simplify generative AI app development, LoftLabs introduces a new way to secure multitenant Kubernetes, and the quote of the week.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Ship it

Down with MCP: As part of Microsoft's 50th birthday party on Friday, GitHub brought two presents for users of Visual Studio Code, one of Microsoft's most important but understated products. All developers using the coding editor will be able to turn on GitHub Copilot "agent mode," which was introduced as a preview in February and "takes Copilot beyond answering a question, instead completing all necessary subtasks across automatically identified or generated files to ensure your primary goal is achieved," the company said in a blog post.

Visual Studio Code also now supports MCP, joining the parade of enterprise tech companies getting behind Anthropic's proposal for making it easier for generative AI apps to exchange data. But GitHub also had some bad news for anyone using Copilot with anything other than OpenAI's GPT-4o model: TechCrunch noted GitHub's new "premium requests" program "imposes rate limits when users switch to AI models other than the base model for tasks such as 'agentic' coding and multi-file edits."

Gateways for gateways: The oversaturation point for MCP-related announcements is probably a week away, but Solo.io entered the chat this week with a new open-source project that promises to make it easier to integrate MCP into software tools. The whole point of MCP is to make it easier to connect applications with large-language models, but getting MCP into those apps for "every application client and agent is error-prone and cumbersome for AI development teams," Solo.io's Keith Babo said in a press release.

MCP Gateway is a new project designed to be used with kgateway, an open-source Kubernetes tool for managing traffic across clusters that was donated to the CNCF by Solo.io earlier this year. The idea is to "provide developers with a single, secure MCP tool registry and access point, regardless of the number of tools used by AI agents and applications," according to Solo.io.

Loft wedge: The European leg of the annual KubeCon world tour hit London this week, and Solo.io wasn't the only startup to introduce something new. LoftLabs has been working on making it easier to use Kubernetes in multitenant situations, and this week it announced vNode, which the company said delivers "stronger isolation of tenant workloads directly at the node layer, without the complexity or overhead of traditional solutions."

Kubernetes was originally designed around the assumption that only one user would manage clusters, but multitenancy "has become an important operational mode as enterprises and cloud providers all offer Kubernetes services for different and sometimes competing clients," according to The New Stack. LoftLabs said vNode was designed for companies or organizations using Kubernetes in highly regulated industries or to run compute-intensive workloads, such as machine learning.

Chevy to the levee: Amazon was a late arrival to the LLM wars, which seems less important these days given the plethora of top-tier models available to anyone thinking about generative AI applications. This week it added to its Nova family of models with Nova Act, which resembles Anthropic's computer-use agent and OpenAI's Operator.

"Nova Act is focused on reliable building blocks that can be composed into more complex workflows," Amazon said in a release. The company also released a software-development kit as a "research preview" that developers can use to experiment with the new model.

Misty water-colored memories: Augment Code is one of many startups trying to carve out a position among the "vibe coders" of our time, who are embracing AI coding assistants and editors at a rapid pace. This week it introduced Augment Agent, which the company said "represents a significant departure from other AI coding tools by focusing on helping developers navigate and modify large, established codebases that span millions of lines of code across multiple repositories," according to VentureBeat.

The agent includes a feature called Memories, "which automatically update as you work with the Agent and persist across conversations to constantly improve the code generated, solve your tasks faster, and match your code style and pattern," Augment said in a blog post. And in keeping with the theme of the week, it also now supports MCP.

Stat of the week

It's one of the most hotly debated questions in enterprise software right now: How many companies are actually using AI agents in production? According to a new survey from PagerDuty, 51% of companies have deployed AI agents somewhere in their environment, and 35% expect to deploy something agentic in the next two years.

Quote of the week

"Between that frustration [with SaaS vendors] and the increased productivity of in-house engineering teams. I think we're moving into — in this classic build-versus-buy (dilemma) — a build era for the enterprise." — Neo4j CEO Emil Eifrem, explaining how AI is changing the way companies think about procuring the software they need to conduct business.

The Runtime roundup

Whatever is left of CISA urged Ivanti VPN customers to patch a critical vulnerability that is being actively exploited, according to CRN.

Ai2 and Google Cloud said they would invest $20 million in the Cancer AI Alliance, which is working on using the pattern-matching magic of AI to fight one of the most deadly groups of diseases on the planet.

Thanks for reading — see you Tuesday!