FenxARC™ is an AI governance platform for models and agents. If you are already calling more than one model, ARC is for you. There is no size you have to reach first. Below: what it governs, where it runs, what a technical review needs, and what you pay for.
ARC is an AI governance platform for models and agents. Its multi-model control plane gives users, applications and agents one governed path to approved models, tools and data, so policy and provider changes do not require each integration to be rebuilt.
If you are already calling more than one model, ARC is for you. There is no size you have to reach first. It was built for developers and technical teams before it was built for a buying committee, and the free tier runs the same engine as everything above it.
ARC can do what a gateway does, including a common request path and routing across destinations. Its role is broader. It governs which models and tools are eligible, how private data is handled, and whether a route runs one model or several steps. It governs the model layer rather than replacing the wider AI stack.
No. The Hosted Platform runs on a zero-access architecture. Queries and responses pass through the router in encrypted form, and FenxLabs does not access, read, inspect or decrypt query content. What FenxLabs processes is the routing metadata a service needs to run and to bill: timestamps, model selections and token counts. No analytics tools, no behavioural tracking and no session recording run inside the platform or the chat app.
Your organisation defines what counts as private. ARC evaluates every governed request and applies the configured handling before it ranks a route. That handling can be to route normally, use a private destination, redact, shield, decompose and recombine, or block the request. Where the handling runs depends on the deployment.
The Hosted Platform runs ARC in FenxLabs infrastructure while you operate the connected model environment. The Managed Service can run inside your infrastructure or be hosted and operated by FenxLabs on your behalf. The Self-hosted Licence runs in your VPC, on your hardware or fully air-gapped, maintained and operated by your team. The policy model is the same in all three.
On the Self-hosted Licence, ARC request handling runs inside your environment. An approved route to an external model or tool still sends that destination the context it needs. Private, internal and air-gapped routes can remain inside your network. Review the hosted and managed boundaries against your own architecture.
No platform can establish compliance for an organisation on its own. ARC can enforce the categories, model and tool access, deployment routes and private-data handling you configure at the model layer. Legal interpretation, policies, risk assessment, documentation, training and controls outside the model layer remain yours.
It is a separate FenxLabs service, not an ARC feature. FenxLabs inventories the AI systems in scope, maps preliminary operator roles and the obligations that may apply, reviews controls and evidence, records the gaps and hands over a prioritised remediation roadmap. It is not legal advice, certification or conformity assessment.
For a small number of stable workloads on one provider, a direct integration is often the simpler answer, and we will say so. The problem appears when model choice, access rules, privacy handling and provider changes are repeated across applications and drift apart. ARC centralises those model-layer decisions while your provider relationships stay direct.
ARC can replace model-selection, routing and policy logic repeated across applications, plus some custom provider-abstraction code. It works alongside your applications, agent logic and identity system, your primary data stores, data pipelines and observability platform, your model-provider contracts, your people and compliance programme and business accountability. Map the exact boundary during evaluation rather than assuming it.
ARC is built to make approved models and tools changeable behind one policy layer, and inference stays a direct relationship between you and your providers at their rates. ARC is still a platform dependency of its own, so evaluate its deployment options, interfaces and exit requirements the way you would any infrastructure component.
Yes. Software vendors use ARC as the routing and policy layer inside a product they sell on. Each of your customers gets their own list, holding the models, tools and rules approved for them, and one request runs inside exactly one list, so one customer never reaches another one’s resources.
ARC is quoted rather than published. The Hosted Platform is billed per routing call and starts with a free tier that needs no card. The Managed Service is scoped and quoted for the engagement. The Self-hosted Licence is one annual fee, priced by the number of human users you licence. Model-inference and third-party tool charges are separate from all three. A quote needs a name, email, optional company name and human-user count, with no meeting.
No. Inference stays a direct relationship between you and your providers, at their rates. FenxLabs charges for ARC through the commercial model you choose, not by marking up what your providers bill you. There is no commercial relationship between FenxLabs and any model vendor.
Start with one real workload. Define its active list, the destinations it may use, its private-data rule, the route you expect and the boundary ARC has to run inside. Use the Hosted Platform for a self-serve start, or talk to an engineer if the deployment boundary or operating model has to be settled first.
ARC is built by FenxLabs, an applied AI lab registered in the Netherlands under KvK 91762782 at Herengracht 320, 1016 CE Amsterdam, Netherlands. The company, the team and general enquiries live at fenxlabs.ai. For a product conversation, book time with an engineer. For an account or a product issue, email support@askarc.app. The entity details, the registration and every policy document are in the ARC legal centre.
ARC is live and commercially deployed. This site publishes no named customers, case studies, benchmark results, certifications or service-level commitments, because none of those are cleared for publication. Ask FenxLabs for the security, reliability, operational and contractual evidence your evaluation needs, and we will tell you what exists.
Thirty minutes with an engineer. Your numbers. A straight answer.