Singulr expands AI governance to endpoints & gateways

Singulr AI has expanded its Agent Pulse product to cover local AI tools on employee devices, AI gateways and Model Context Protocol environments, broadening its governance reach beyond browser-based AI services.

The expansion adds oversight for developer tools and desktop applications, including Claude Code, Cursor, Cowork and ChatGPT Desktop. It also extends controls to AI gateways such as LiteLLM and to internally built agentic applications through the Singulr software development kit.

The change reflects a shift in how companies are using artificial intelligence at work. Rather than relying mainly on browser-based assistants, staff are increasingly running AI tools directly on their machines. Developers are using coding assistants on workstations, while platform teams are deploying gateways that connect models, data and internal systems.

At the same time, AI agents are gaining access to enterprise applications and data through MCP servers, a growing standard for connecting models to tools and information sources. In some cases, employees are linking untrusted servers directly to local AI tools, creating security and governance issues that existing controls were not built to manage.

That trend has left many organisations with fragmented oversight across endpoint, network, cloud and identity systems. The latest expansion is intended to give security, IT, privacy and risk teams a more consistent way to apply policy across those environments.

Broader coverage

The updated version of Agent Pulse includes an endpoint sensor designed to monitor AI activity on developer workstations and desktop applications. It also adds content inspection and guardrails for AI gateways, runtime controls for internally built agentic software, threat analysis for MCP environments, and analytics on Microsoft 365 and Gemini usage and access.

These features are backed by the Singulr Pulse risk intelligence engine, which maps more than 2 million signatures of AI assets and risks, including MCP servers. Singulr described MCP as one of the more exposed parts of the emerging agent-based software stack.

"The perimeter has moved," said Shiv Agarwal, Co-Founder and Chief Executive Officer, Singulr AI. "AI agents are running on employee endpoint devices, interacting with enterprise tools and accessing data through MCP. Organisations moving fast with agentic AI need governance that keeps pace, and now they have it. Agent Pulse is the only runtime control plane that follows AI wherever it operates, from the browser to the endpoint to the agentic workflow, while preserving the context and policy alignment enterprises need."

The announcement highlights a wider challenge for companies trying to track AI use as tools spread beyond centrally managed cloud applications. Governance products first focused on browser plug-ins and approved software lists, but security teams are now also confronting locally installed assistants, coding tools and agent frameworks that interact more directly with internal systems.

Customer view

Trace3, an IT consulting and services company focused on AI, data, cloud and cybersecurity, is among the users of Singulr's software. It uses Singulr to help govern enterprise AI adoption.

"Our clients are accelerating their adoption of agentic AI, and we've experienced that same shift internally," said Bryan Kissinger, PhD, Senior Vice President of Security Solutions and Networking and Chief Information Security Officer at Trace3. "The challenge isn't simply discovering AI applications anymore; it's understanding what autonomous agents are doing, what systems they're connected to, and ensuring they operate within policy. As a Singulr customer, we've seen the value of having a unified control plane that provides visibility, governance, and verifiable assurance for AI agents as they're introduced into the enterprise. It's an important capability for organizations that want to innovate without sacrificing security or trust."

Singulr launched Agent Pulse earlier this year. With the latest expansion, it is positioning the product around runtime governance across browser-based tools, endpoints, gateways and agent platforms, rather than only monitoring use of public AI services.

Its products are also now available through AWS Marketplace, giving customers another way to buy the software through existing cloud purchasing arrangements.

The expanded controls now span cloud agentic platforms, browser-based AI, local AI desktop applications, coding assistants, MCP and AI gateways.