Skip to content

Migration guidesFrom Requesty

From Requesty, onto your own provider accounts

Requesty is a hosted router: one key, its own provider access paid from prepaid credits, and your own keys for six providers if you want them. Its routing policies, spend limits and request metadata map closely to INFRO. The difference is whose account serves the request: on INFRO it is always yours.

Independent; hosted router with its own provider access on credits, and BYOK for six providers

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

Independent. Its blog's latest company news is a seed round in September 2025, and the blog is active through October 2026. We found no acquisition or shutdown.

Whose account serves the request

By default Requesty's own provider access, from prepaid credits. BYOK covers OpenAI, Anthropic, Google AI Studio, Vertex AI, Mistral and xAI, and a policy can mix its keys and yours.

What moves

Base URL and key, policy/<name> to a route, the requesty object to metadata, auto_cache to the provider's own caching. HTTP-Referer and X-Title come out.

If you want model access without an account at each provider, Requesty solves that and INFRO does not.

  • You call models from providers you hold no account with. INFRO only runs on accounts you connect.
  • You use latency routing between providers. INFRO's routes are priority or weighted.
  • You rely on its prompt library or auto caching that inserts cache breakpoints for you. INFRO has neither.
  • You need its EU router endpoint. INFRO's residency is on the roadmap.

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

 RequestyINFRO
Base URLhttps://router.requesty.ai/v1https://api.infro.io/v1
KeyA Requesty key; provider keys optional.An INFRO gateway key; your provider accounts are connections.
Model idgpt-5.6-terra, openai/gpt-5.6-terra, policy/<name>.openai/gpt-5.6-terra, or a route name in place of policy/<name>.
Body fieldsrequesty { tags, user_id, trace_id, extra, auto_cache }.metadata { user_id, session_id, tags }; X-INFRO-Trace-Id.

Before · Requesty

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://router.requesty.ai/v1",
  apiKey: process.env.REQUESTY_API_KEY,
});

await client.chat.completions.create({
  model: "policy/chat-default",
  messages: [{ role: "user", content: "Hello" }],
  // @ts-expect-error Requesty extension
  requesty: { user_id: "user_123", tags: ["chat"] },
});

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: "chat-default",
  messages: [{ role: "user", content: "Hello" }],
  // @ts-expect-error INFRO extension
  metadata: { user_id: "user_123", tags: { feature: "chat" } },
});

Policies become routes

A fallback policy is a priority route; a load-balancing policy a weighted route; per-model retries the route's failure rules. Name the route after the policy and send it as model.

Tags become key and value

Requesty's tags are a list; INFRO's metadata.tags is a map, filtered on the request log as tag.<key>.

Models need an account

List what Requesty served from its own access, and connect a provider account for each one you keep using.

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

RequestyWhat it does thereIn INFRO
Requesty key and creditsdifferentOne key over Requesty's provider access, paid from prepaid credits.A gateway key over your own connections; each provider bills you directly.
Fallback and load-balancing policiessame shapeOrdered or weighted model chains with retries.Routes with priority or weighted targets and failure rules per kind.
Latency routingnothingPicks the fastest provider.Nothing.
Spend limits and alertssame shapeMonthly cap per key or service account; spending alerts.Budgets per organization, project, member, key and end customer, 402 before the call, alert thresholds.
Guardrailssame shapeScan requests for PII and credentials.Pattern rules that block, redact or flag PII, keys and your own terms; decision-model checks.
Request metadata and cost trackingsame shapeTags, user and trace ids; cost per request.metadata on every request; estimated provider cost, unpriced when no rate is known.
Auto cachingdifferentInserts provider prompt-cache breakpoints.Nothing that inserts breakpoints. An opt-in exact-match response cache.
Prompt librarynothingStored prompts.Nothing.
MCP gatewaysame shapeMCP servers behind the router.An MCP gateway for your own MCP servers.

Our reading of Requesty’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

  • Serve models from providers you hold no account with.
  • Route by latency.
  • Insert prompt-cache breakpoints or keep a prompt library.
  • Pin traffic to the EU.

What does not come across

  • Request history and analytics. There is no importer.
  • Credits on your Requesty account.
  • Policies and limits as objects. Recreate them as routes and budgets.

Questions

What replaces `policy/<name>`?
An INFRO route with the same name, holding the same models in order or by weight.
Does INFRO sell model access like Requesty's credits?
No. INFRO runs on your own provider accounts, which bill you directly. INFRO is a subscription with 0% markup on model spend.
Can I keep my Requesty metadata?
Yes, renamed: user_id stays user_id under metadata, trace_id becomes X-INFRO-Trace-Id, tags become a key and value map.

Move with us, or on your own

The migration program maps your Requesty 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 RequestyRead the quickstart

Sources

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