Agate

Overview

Governance framework for AI agent actions | SaaS · Hybrid | AI · Security · Governance

Companies are developing AI agents capable of connecting to their applications, APIs and data. When these agents begin to act autonomously within the Information System, a new question arises: how can their actions be governed without hindering their roll-out?

Aggate introduces a new layer of infrastructure between AI agents and the organisation’s resources. Each agent has an identity, and every action can be assessed based on its intent, permissions and context, and then authorised or blocked before execution.

The platform centralises agent governance and enables teams to deploy their use cases at scale, without having to recreate security mechanisms for each project.

How to use

Cloud or hybrid deployment

Aggate adapts to organisations’ infrastructure and sovereignty constraints with two deployment options.

In managed cloud mode, the platform is hosted on the Cloud Temple infrastructure in France, with operation and maintenance provided by Aggate.

In hybrid mode, certain components can be deployed within the customer’s environment to meet their requirements in terms of architecture, security and data processing.

Connecting agents

The administrator registers agents and their resources via the Aggate console. The platform can integrate with the corporate directory to link actions to the agent and, where applicable, to the user on whose behalf the agent is acting.

Agents connect to a shared access point that provides access to the MCP tool servers and authorised language model providers. The gateway manages the credentials required to access the resources, without exposing the services’ secrets directly to the agents.

With every call, Aggate authenticates the agent, checks the relevant authorisations and routes the request to the appropriate tool or model.

Authorisation policies and simulation

The platform and security teams define policies that take into account agents, users, tools and the execution context. Actions that are not explicitly authorised are denied by default.

The console allows you to simulate authorisation decisions before going live, view actual access rights, and amend, suspend or revoke authorisations where necessary.

Traceability and monitoring

Every decision is logged to maintain visibility over staff actions, the resources utilised and the policies applied. These events can be utilised by operations and security teams, in particular via their monitoring tools and SIEM systems.

Start-up support

Aggate supports organisations in rolling out their first use case, from identifying staff and resources to connecting them, defining policies and commissioning the system. Technical and security teams are involved at every stage to ensure a reusable governance framework is in place for future agents.

Support

Support and guidance are provided directly by Aggate. To discuss an initial use case or the deployment process, contact the partner via the Cloud Temple Marketplace or visit aggate.ai.