The device layer of the Guardian Agent Platform.
Guardian agents that find shadow AI, connect tools to their accounts and access, and enforce your policy on supported AI requests where the work happens.
Follow a supported AI interaction from local discovery to an enforceable decision. Scroll to explore, or select a step.
Read supported local signals to find AI apps, coding agents, local models and MCP connections. Discovery needs no per-tool API integration, traffic interception or execution of the tools it finds.
For supported enforcement paths, a local TLS gateway reads the request on the device and re-encrypts forwarded traffic. Policy checks use deterministic rules and trained classifiers, with no language model required.
Let approved work continue. Warn, redact or block supported interactions when a rule applies. The action follows the policy your organisation publishes.
Request controls are distinct from stopping an application or removing it. App removal needs an administrator; coverage depends on the tool and operating system.
Build a versioned, redacted record on the endpoint. Evidence is saved locally, then delivered through an encrypted outbound route to the platform.
Framework mappings support control reviews; they do not establish certification or guarantee compliance.
Follow the evidence ↗The agent observes AI use and applies supported controls on the endpoint. Your teams manage policy and investigate the evidence in one console.
Discover supported AI, evaluate local interactions and produce redacted evidence. Local discovery can continue offline.
Attribute available usage and spend to tools, people and projects. Publish policies, assign findings and use the evidence to review AI value alongside business outcomes.
Validate the signals and controls on your own workflows, then use your existing device management to roll out the agent.
Install the agent on a supported device. Enrol with a claim code that an authorised user claims in the console.
Review the first inventory, confirm data collection and test policy actions against the tools your people use.
Distribute through Intune, Jamf or SCCM. Unattended enrolment exchanges an organisation fleet secret for a device credential.
Choose the current package and configuration for your environment.
View installation options ↗Discovery, request inspection and evidence reporting serve different purposes. Here is how the data is handled at each step.
Discovery reports that a credential exists. Secret values are removed on the endpoint before evidence is built.
Discovery identifies repository context and bounded file inventory. It does not collect source-code contents.
Prompt, response and transcript collection is off by default. Administrator-enabled collection is distinct from local inspection for enforcement.
Local decisions use deterministic rules and trained classifiers. Your request is not sent to a language model to make that decision.
A rollout is more than an install. Reporting state, policy versions and evidence delivery need to remain visible.
Publish a new policy version from the platform. Keep earlier decisions tied to the version that produced them.
One failed collection check does not stop the scan. Missing evidence is not presented as a pass.
Local discovery continues offline and evidence is saved before delivery. The console shows the last reported state, including paused or silent devices.
Confirm the deployment-specific details with your team before expanding to the fleet.
Measure CPU, memory, scan time and request latency on representative devices and workflows. Resource use depends on your configuration and enabled controls; agree acceptance criteria before expanding.
Offline discovery and local evidence storage are covered above. For request enforcement, confirm cached-policy behaviour, fail-open or fail-closed settings and queue limits for the paths you enable.
The separate open-source Surface tool lets you inspect AI activity on a machine. It is not the managed Layer 1 Agent or the full platform. Explore open-source Surface ↗.
See the Guardian Agent Platform on your own fleet. Find every AI tool, understand what it reaches and costs, and apply the controls your team needs.