FenxARC™ is an AI governance platform for models and agents. It gives users, applications and agents one governed path to approved models, tools and data, enforcing access, privacy, routing and spending policy before requests run.
Free tier on the Hosted Platform. No card required.
anthropic-us-west.primaryAnthropic · OpenAI · Google AI
equinix-am5.sglang-03SGLang · Ollama · llama.cpp
bedrock-eu-central.prodAmazon Bedrock · Google Vertex · Microsoft Foundry
nebius-eu-north-h100.poolNebius · Scaleway · OVHcloud
fireworks-us-virginia.slsFireworks · Together · Baseten
portkey-eu.guardrailPortkey · Kong · Cloudflare Gateway
Set approved models, tools, data access, privacy and spending rules in one place. ARC applies the assigned boundary to each user, application, workspace and agent before every request runs.
Give each user, application and agent only the approved models, tools and data access its work needs.
Apply budgets and cost preferences by person, team or use case. Cost can reorder approved routes but never relax access, privacy or capability.
Add providers, replace models and update rules without rebuilding every application.
ARC resolves the caller and its approved boundary, applies access and customer-defined privacy rules, then ranks only permitted routes. Cost can change the order, never the permissions.
Policy determines which models, tools and route options are permitted. ARC ranks only those permitted options.
A cost objective reorders routes that already meet your privacy policy. It cannot reach one that does not.
Unauthorised requests are blocked before any tokens are consumed.
When a provider fails, the request takes the next path you set. A local rule-based path scores requests without calling out to anything, so routing still happens when the systems it routes to are down.
Use budgets and routing preferences to reduce spend on eligible work without relaxing access, privacy or capability requirements.
Modelled at the balanced setting
70%potential AI spend reduction
ARC first filters for access, privacy and capability. It only optimises cost among routes that already meet the request.
Modelled against routing the same mixed workload entirely to frontier models. Your result depends on your providers, traffic and workload mix.
Choose the deployment model that fits your infrastructure, security requirements and operating capacity. Start hosted, use the Managed Service, or run ARC yourself with the Self-hosted Licence.
Start free on a FenxLabs-hosted ARC service. Connect your endpoints, set policy and operate the wider model environment yourself. Nothing to deploy.
Start on the Hosted PlatformHave FenxLabs operate ARC and the agreed AI environment inside your infrastructure or host and operate it on your behalf.
Explore the Managed ServiceRun and maintain ARC inside your VPC, on your hardware or fully air-gapped. Your team owns day-to-day operation, with standard FenxLabs support.
Explore the Self-hosted LicenceFenxLabs can assess the workload, design and integrate the architecture, adapt models and prepare governance. Ongoing responsibility belongs to the Managed Service above.
How an engagement runsThe API speaks OpenAI. Point it at a new base URL and your client is unchanged.
Omit the model and ARC scores the request and routes it. Name a model and the request goes there instead of being routed. Pinning is bounded by the model lists assigned to the key in use.
from openai import OpenAI
client = OpenAI(
base_url="https://api.askarc.app/api/v1",
api_key="fxk_your_api_key_here",
)
# No model named, so ARC routes the request.
response = client.chat.completions.create(messages=messages)ARC is built by FenxLabs in Amsterdam and is live in commercial deployments.
Start free with one integration and your own endpoints. Talk to an engineer if deployment, data boundaries or operating responsibility need to be settled first.