Gravitee has launched Gamma, an agent management platform. Research commissioned by the company found that British businesses have deployed 713,130 AI agents.
A survey of 750 chief technology officers and vice presidents of engineering at major enterprises also found that the number of AI agents used by UK firms has more than doubled from 250,000 in January. Gravitee noted that the total now exceeds the combined number of GPs, police officers and solicitors in Britain.
AI agents are software systems designed to carry out tasks with limited or no direct human intervention. Their growing use in large companies points to a shift from narrower chatbot-style tools to more autonomous systems that can access software, trigger workflows and interact with other digital services.
With Gamma, Gravitee is targeting that shift with a platform designed to govern how such agents operate inside organisations. The product combines an AI gateway, a catalogue for model context protocol assets and an authorisation layer intended to control access to tools, systems and other agents.
Rising use
The UK figures form part of a wider Gravitee estimate that the global population of AI agents is now larger than Denmark's population. While the numbers are based on polling rather than direct measurement, they suggest that large enterprises are moving quickly to introduce autonomous software workers into day-to-day operations.
That expansion is likely to sharpen questions around oversight, security and accountability. As more agents are allowed to call models, access internal tools and trigger actions across business systems, companies face the task of setting rules on what each agent can do and tracking how those actions unfold.
Gamma is designed to sit in that control layer. The platform combines LLM Proxy, MCP Proxy and A2A Proxy functions in one gateway, giving organisations a single place to govern model calls, tool usage and agent-to-agent interactions.
The platform also includes a catalogue of MCP servers, tools, skills, resources and agents. It can import from existing MCP registries and operate as a registry itself, allowing businesses to organise and expose AI-related assets from one system.
Control layer
Another part of the product is MCP Studio, which allows engineering and platform teams to create new MCP servers from existing tools, legacy systems, application programming interfaces and event streams. The approach is intended to help companies connect older technology estates to newer AI systems without rebuilding everything from scratch.
At the centre of the platform is an authorisation engine based on Cedar, the open source policy language originally developed by Amazon. Gravitee says this allows organisations to define detailed policies on catalogued assets and enforce them through its gateways, with the aim of making zero-trust controls standard from the outset.
The company has also added what it calls agent identities, using cryptographic workload identity for AI agents. These controls include on-behalf-of token exchange and integration with SPIFFE and SPIRE, which are used to establish trusted software identities in distributed systems.
Lineage, another feature in the release, is designed to provide an audit trail across agent activity, including links between agents, tools and language models. Gravitee says this should allow companies to trace behaviour, assign costs and record policy decisions at each stage of an interaction.
Rory Blundell, Chief Executive Officer and Co-founder of Gravitee, linked the product launch to the broader growth of autonomous systems inside companies.
"The number of agents in the workforce is soaring, but they're also getting ever more powerful. Frontier models are getting more capable every month, and robotics is maturing fast. Soon the same models running in enterprise software will be running in machines on factory floors and warehouses. As we give LLMs arms and legs to perform actions, it is essential that we also give them a central nervous system. Gamma is the control layer that makes the agentic future safe," said Blundell.
Security focus
The platform is aimed at two broad groups: internal teams managing AI operations and organisations building AI-based products for customers or partners. In the first case, the focus is on routing model traffic, securing access to external MCP servers and monitoring costs. In the second, the platform is intended to help businesses expose internal context and services through governed MCP endpoints.
Gravitee, which says it has a valuation of more than USD $300 million, has built its business around managing and securing digital infrastructure, including application programming interfaces and event streams. Gamma extends that strategy into AI management at a time when governance tools are becoming a more crowded part of the enterprise software market.
Linus Håkansson, Chief Product and Technology Officer at Gravitee, said the product was designed to address both speed and oversight in enterprise AI deployment.
"Gamma is built on the belief that security and speed are not a trade-off. With a catalog of every asset, authorisation policies defined against those assets, and enforcement at every gateway, you can build and expose AI capabilities at pace and know exactly what's happening at every step," said Håkansson.