Presented by Fabrix.ai
In the latest VB Spotlight, Fabrix.ai explores how enterprises can govern vibe coding as it spreads beyond traditional developers, with Governed VibeOps and Argos models designed to address reliability, control and cost.
View on demand here.
Reliability has overtaken token cost as the top AI agent metric for enterprises, according to VentureBeat Pulse survey data. At the same time, vibe coding is spreading beyond traditional development teams, with marketing and sales staff increasingly committing code alongside developers.
“The industry doesn’t have a vibe coding problem, they have a vibe governance problem,” Manjrekar explained to fellow speakers, Rached Blili, distinguished engineer at Fabrix.ai, Rob Strechay, lead analyst, VB Intelligence, and host Sam Witteveen, VentureBeat Senior Technology Contributor.
Fabrix recently launched Governed VibeOps, which lets IT operations staff specify dashboards, AI agents, agentic workflows, and automation pipelines in natural language. A central platform inspects the generated code and meters token spend, and the company has had production customers on the capability since almost the beginning of this year.
Centralized governance can accelerate rather than constrain vibe coding
Vibe coding has given users the courage to design entirely new ways of looking at data or interacting with it, Blili said, and that should be encouraged.
“When you implement governance properly, you’re not going to slow down those creative instincts. You may actually amplify them,” he said.
But the requirements change when an application needs to operate inside an enterprise. It may need to read production data, write changes back, trigger workflows, and respect whoever holds permission to do those things, requirements that are difficult to meet on a laptop with a personal coding agent. A platform that already holds the connections, the permissions model, and the map of where data lives means the agent inherits access control, auditability, traceability, cost metering, and evaluation without the user assembling any of it. Development can get faster, and those activities centralize onto the platform.
“If you don’t centralize or govern VibeOps or vibe coding at all, you could have 500 people developing exactly the same application on their own computers,” Blili added. “Whereas if you were providing some sort of place for that to go, one person could develop the application, share it with other people, they could pick it up and maybe improve on it and share those improvements with other people, and you have a lot more efficiency.”
But platform teams need to watch individual sandboxes, inspect the code coming out of them, and confirm that nobody is consuming tokens out of contract.
The VB Pulse survey found 63% of polled organizations were either running a governed semantic or context layer in production or piloting or building one, with another 20% actively evaluating one. Strechay characterized that as governance moving out of policy documents and into the execution path.
“DevOps really governs the delivery pipeline,” he said. “What I love about VibeOps is it’s aiming to govern the intent, the context, and the agent decisions that flow through it.”
Three Argos models handle platform semantics, AIOps and vulnerability exposure
Governed VibeOps runs as a capability of the Fabrix.ai agentic platform, which is used at large enterprises for operational intelligence across IT, network, and security silos. Users can bring their own coding agent or use Fabrix.ai’s built-in AI Coding Assistant with LLMRouter.
Underneath that is the platform infrastructure to which the coding agent can offload much of the detail work. The platform is built on a Universal MCP server, which enables connectivity to hundreds of multi-vendor/domain datastores, a context engine which ensures context purity and token maxxing with a memory architecture and a harness with Observability, FinOps and Evaluations. Agentic data federation creates a living ontology that maps what data exists across the enterprise and where to find it, giving agents the context to locate data without customers first ingesting everything into Fabrix.ai. Universal connectivity will expose the vendor’s integration catalog to anything built through VibeOps.
“We recently did a major facelift on a FinOps dashboard that we built using Codex,” Blili said. “We just gave it a specification, and it went and rejiggered the whole thing and massively upgraded the functionality and the usability of the system.”
A key part of that governance layer is a family of three small language models under the Argos name, which Fabrix.ai trains on a customer’s own environment and data. Argos VX carries the platform semantics, generating vibe-coded applications, AI agents, and the vendor’s own data pipelines. Argos AIOps holds correlation policies and reads the customer’s environment and telemetry. Argos VE scores vulnerability exposure, answering what a newly disclosed flaw means for a given estate and how far the blast radius extends.
“You don’t need a frontier or foundational model for every task you’re trying to accomplish in your workflow,” Manjrekar explained.
According to Fabrix.ai, the models range from roughly 4 billion to 8 billion parameters. Because they are trained on the customer’s environment, Manjrekar explained, data stays put while the models gain accuracy and avoid the latency of going over the wire. Fabrix.ai’s own pipelines consume no tokens, he added, moving a large share of data management work off the meter. The distributed compute harness can also be accessed by other agentic frameworks through an MCP endpoint.
For customers, Manjrekar described a progression from initial productivity gains to tool consolidation and, eventually, a blurring of traditional operational disciplines.
“Right now, SecOps and ITOps are essentially working off of the same set of data,” he said. “Over a period with VibeOps, you would see this collapsing around data.”
CFOs want business outcomes, and FinOps shows them a monthly bill
Strechay argued that FinOps has become disconnected from the work, reduced to reading a monthly bill at the end of it, while tokenomics has to start by connecting the agent, workflow and business outcome. That becomes increasingly important as organizations deploy code faster and traditional QA gates and DevOps pipelines struggle to scale when the developer population jumps by an order of magnitude.
“CFOs don’t understand tokens, and they don’t want to understand tokens,” he explained. “They want to understand business outcomes from development. And if you’re having these many more developers, they want to know what their investment is gaining them.”
To learn more about how enterprises govern vibe coding as development expands beyond traditional developers, control token spend and keep agents reliable, don't miss this VB Spotlight.
Agenda
Why extending vibe coding into enterprise operations requires a governed approach
How to connect directly into an agentic AI platform to build bespoke apps and dashboards
Managing security, support, and tokenomics in a vibe coding organization
Building, deploying, observing, and optimizing agentic applications, agents, and dashboards
Presenters
Shailesh Manjrekar, Chief Marketing Officer, AI Strategy, Fabrix.ai
Rached Blili, Distinguished Engineer, Fabrix.ai
Rob Strechay, Lead Analyst, VB Intelligence
Sam Witteveen, Sr. Technology Contributor, VentureBeat (Moderator)