Skip to content

Migration guidesFrom Orq.ai

From Orq.ai's AI Gateway

Orq's AI Gateway is OpenAI-compatible under /v3/router, takes provider/model ids, and runs either on Orq's credits or on provider keys you connect. The gateway moves to INFRO in four edits. The AI Studio around it (prompts, evaluators, experiments, knowledge bases, agents) does not move, and the case for staying comes first.

Amsterdam-built; AI Gateway on Orq credits or your own keys; EU SaaS, or VPC and on-premises on Enterprise

INFRO accounts are self-serve. The catalog, the docs, and the API contract are public, the gateway is deployed, and anything beyond v1 is marked Later. Plans and trial

Where things stand

Company

Amsterdam-built. It describes itself as an EU-built control plane for enterprise AI, and lists an agent-driven Software Factory in early access beside the gateway and AI Studio.

Whose account serves the request

Either. On Orq's credits, Orq buys the inference; with your own keys, billing stays with your provider. INFRO has only the second arrangement: before moving traffic that runs on Orq's credits, open an account with that provider and connect its key.

Where it runs

EU SaaS on Google Cloud by default; on Enterprise, a VPC on AWS or Azure, a single Helm chart on-premises, or air-gapped.

Orq is a platform for building AI applications, with a gateway in it. Stay if you use the platform, or need it inside your own perimeter.

  • Your prompts, evaluators, datasets, experiments or knowledge bases live in AI Studio, or your agents run as Orq agents (agent/<name> as the model). INFRO has none of these.
  • Processing has to stay in the EU, or the gateway has to run in your VPC, on-premises or air-gapped. INFRO is hosted; its gateway runs in Frankfurt, which is not a residency control, and self-hosting is an enterprise agreement.
  • You route on Orq's credits and do not want an account with each provider. INFRO never buys inference for you.
  • You rely on latency-based or round-robin load balancing, or on budgets per provider and per model.
  • You use model-judged, Python or external guardrails such as Lakera.

The edits, one row each. Everything not listed here stays as it is.

 Orq.aiINFRO
Base URLhttps://my.orq.ai/v3/router, or your own host on VPC and on-premises.https://api.infro.io/v1
KeyAn Orq API key.An INFRO gateway key per service; members sign in to the console separately.
Model idopenai/gpt-5.6-terra, enabled on the Models page.Usually unchanged: INFRO also takes provider/model. Check each id on /models, or send a route's name.
Fields and headersfallbacks: [{ model }], retry, cache: { type, ttl }, identity or X-ORQ-IDENTITY-ID, thread, metadata and X-ORQ-METADATA-*.fallbacks as model ids (up to 5), retries on a route, cache: { ttl_seconds, namespace }, metadata { user_id, session_id, tags }.

Before · Orq.ai

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://my.orq.ai/v3/router",
  apiKey: process.env.ORQ_API_KEY,
  defaultHeaders: { "X-ORQ-IDENTITY-ID": "user_123" },
});

await client.chat.completions.create({
  model: "openai/gpt-5.6-terra",
  messages: [{ role: "user", content: "Hello" }],
  // @ts-expect-error Orq extensions
  fallbacks: [{ model: "anthropic/claude-sonnet-5.5" }],
  retry: { count: 3 },
});

After · INFRO

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.infro.io/v1",
  apiKey: process.env.INFRO_API_KEY,
});

await client.chat.completions.create({
  model: "openai/gpt-5.6-terra",
  messages: [{ role: "user", content: "Hello" }],
  // @ts-expect-error INFRO extension
  fallbacks: ["anthropic/claude-sonnet-5.5"],
  metadata: { user_id: "user_123" },
});

Identities become `metadata.user_id`

Orq's identity object, or the X-ORQ-IDENTITY-ID header (which attributes a request only to an identity that already exists), becomes INFRO's metadata.user_id: your own opaque id of up to 128 characters, or the X-INFRO-User-Id header. There is no record to create first, and INFRO stores no name or email for it.

Fallbacks are plain model ids

Orq's fallbacks are objects with optional fallback_on_codes; INFRO's are a list of model ids, tried after every connection for model has failed. Retry counts, status codes and the order across your providers belong on a route's failure rules.

Routing rules become routes

An Orq routing rule rewrites a request whose model matches; an INFRO route is a name your application calls. Conditions on headers and metadata become the route's if-rules, and fallback or weighted load balancers become priority or weighted targets.

Orq.ai's objects against INFRO's, from Orq.ai's public documentation read on 19 October 2026 and checked against what the gateway implements. Three rows say INFRO has nothing.

Orq.aiWhat it does thereIn INFRO
BYOK provider keyssame shapeProvider keys connected in the AI Gateway, several per provider.Provider connections, several keys per provider.
Orq creditsnothingOrq buys the inference from one balance across providers.Nothing. INFRO holds no provider accounts; connect your own.
API keyssame shapeKeys for the gateway, with budgets and gateway scopes per key.Gateway keys per service with allowlists, IP ranges, environments and spend limits.
Fallbacks and retriesdifferentfallbacks and retry per request, with status codes.fallbacks per request; retries and status codes on a route's failure rules.
Routing rules and load balancingdifferentCEL rules on model, header, identity, metadata or project; fallback, latency, weighted or round-robin.Routes with if-rules on metadata, headers and size, and priority or weighted targets. No latency or round-robin strategy.
Cachesame shapeExact-match cache per request, with a TTL.An opt-in, exact-match cache for chat completions, with a namespace and TTL.
BudgetsdifferentWorkspace, project, identity, API key, provider and model; cost, tokens and requests per minute; 429.Organization, project, member, key and end customer; dollar ceilings that refuse with 402 before the call. No provider or model budgets, and no token-count windows.
IdentitiesdifferentEnd users as records with name, email and tags; budgets per identity.metadata.user_id on each request; end-customer caps from the Team plan.
GuardrailsdifferentPII, secret detection, model-judged, Python and external guardrails, on input and output.Pattern rules and prompt classifiers on chat, per organization, project or key; Business plan.
MCP Gatewaysame shapeUpstream MCP servers behind one address, OAuth sign-in, tool exposure per server.An MCP gateway with approved and pinned tools and grants per key; Business plan.
Prompts, evaluators, experimentsnothingAI Studio, usable from router requests.Nothing.
Knowledge bases and agentsnothingManaged retrieval and agents, callable through the gateway.Nothing. INFRO routes your own agents' model calls.

Our reading of Orq.ai’s published documentation, not a trial or a benchmark. If we have read something wrong, tell us and we will fix the row.

What INFRO does not do

  • Buy inference for you on a platform balance.
  • Prompt management, evaluators, experiments, knowledge bases or managed agents.
  • Latency-based or round-robin load balancing.
  • Budgets per provider or per model, or token-count budgets.
  • Model-judged, Python or external guardrails.
  • EU-only processing, or running in your VPC without an enterprise agreement.

What does not come across

  • Traces, threads and identity records. There is no importer.
  • Routing rules, budgets and guardrail rules as objects. Recreate them.
  • Prompts, datasets, evaluators and knowledge bases, which stay in AI Studio.

Questions

Is Orq.ai BYOK too?
It can be. Orq runs either on its own credits or on your provider keys; INFRO runs only on your provider accounts, and takes 0% of model spend.
Can I keep AI Studio and move only the gateway?
Yes, for the calls your application makes to models directly: point those at INFRO. Calls that use Orq's managed prompts, agents or knowledge bases go through Orq's router, and stay there.
What does INFRO cost?
A flat subscription by plan, with a 14-day trial. Your providers bill you for model use directly.

Move with us, or on your own

The migration program maps your Orq.ai setup to INFRO connections, routes and budgets on a call and moves the first service with you. Your providers keep billing you directly; INFRO is a subscription and adds 0% to model spend.

Start 14-day trial Book a migration sessionINFRO vs Orq.aiRead the quickstart

Sources

Every statement on this page about Orq.ai comes from one of these, read on 19 October 2026. Nothing is inferred from a trial, a benchmark, or a conversation.