An AI gateway for guardrails, cost visibility and tracing.
When AI applications use different tools and providers, security and engineering teams need a defined control point and a way to inspect usage. Linkfields AI Governance packages an AI gateway and governance layer built on Portkey + Opik.
Portkey and Opik are third-party components, not platforms owned by Linkfields. The Linkfields packaging, integration and governance configuration must be scoped alongside the underlying platform licences and operating responsibilities.
This product sits in Cyber Security. It is distinct from AI strategy & governance, which addresses use-case priorities, ownership and organisational governance.
For security control design and GRC support, explore AI security & GRC. A gateway alone does not establish legal or regulatory compliance.
The product offer covers guardrails, cost and tracing. Start with the controls and evidence your AI workload needs, then establish the supported configuration.
Translate AI-use rules into a control checklist: what should be permitted, restricted or reviewed. Demonstrate the relevant gateway behaviour before approving the configuration.
Use cost visibility to assess the spend associated with the agreed AI workload. Confirm the available usage views and attribution detail rather than assume universal budget enforcement.
Inspect the path of an AI interaction to understand how the workload behaves. Agree which trace evidence is needed and who may inspect prompt and response content.
Evaluation workflow, not a production architecture: agree the application, gateway, model/provider and tracing paths for one workload before expanding coverage.
Identify the application, AI tools, model providers and data types. Record trust boundaries and the current operating owners.
AI estate and responsibility map
Confirm what Portkey + Opik provide and what Linkfields configures or integrates. Check provider coverage and third-party terms.
Component and licence boundary
Use agreed sample interactions to inspect guardrail behaviour, usage visibility and trace evidence. Include failure-handling questions.
Policy-test and evidence record
Set access, secrets, retention and incident responsibilities. Accept the design only against the controls actually demonstrated.
Operating and acceptance checklist
Example engagement pattern, not a client case study. An enterprise team wants to introduce a knowledge assistant but first needs visibility of its AI interactions. Start with one workload, agreed sample prompts and a named control owner.
The evaluation produces an application-to-gateway map, a supported-control checklist, sample usage and trace evidence, and a responsibility matrix. Organisational policy and security review remain separate from the technical gateway configuration.
For the wider evidence and remediation programme, explore Regulated Cyber Readiness.
Available by licence, subscription or SaaS, with or without Linkfields services. A walkthrough should establish the Linkfields scope, third-party terms and supported estate before procurement.
Let’s connect
