Skip to content

AI infrastructure control plane

Your providers. Your keys. One control plane.

Connect the AI provider accounts you already pay for, and get one place to route across them, fail over automatically, trace every request and control spend. Your contracts and your rates stay yours — INFRO takes 0% of what you spend on models.

Get early accessSee the console
Operations
Requests
2.23M
+12%
Success
99.78%
+0.05%
Spend
$6,142
$2,880 saved

Request volume

Last 24 hours

Normal

Active routes

Automatic allocation

  • Primary

    GPT-5.6 Terra

    64%

  • Standby

    Claude Sonnet 5

    25%

  • Overflow

    Gemini 3.7 Flash

    11%
INFRO console showing traffic health, request volume, success rate, estimated provider spend, connection health, and automatic failover.

When a model call becomes infrastructure.

One call is easy. INFRO becomes useful when provider complexity, outages, and spend start competing with the product you meant to build.

A dense stream of request pulses narrowing through a central chamber into fewer, steadier channels, with a ledger plane beneath recording each one.

  1. 01A second provider adds a second integration

    SDKs, request shapes, and fallback logic start leaking into your code.

  2. 02Model spend starts squeezing margin

    A monthly total from each provider no longer tells you which request, customer, or feature costs what.

  3. 03A provider incident becomes your incident

    When the upstream fails, your customers feel it first.

  4. 04More people need their own guardrails

    Teams, environments, and customers need separate keys, budgets, and model policies.

The economic case

Your providers bill you for AI. INFRO bills you for control.

Two costs that never move together. Model spend stays with your providers at your own rates, and the control plane is $99 a month, flat — 0% markup on model spend.

Your providers bill you, at your rates
Requests run on your own provider accounts, so your contracts, committed-spend discounts and enterprise agreements keep working exactly as they do today. INFRO takes 0% of model spend.
One flat cost for the control plane
Routing, failover, request logs, cost attribution, budgets, roles and the audit log are the subscription. It does not move when your token usage does.
The estimate is yours to correct
INFRO estimates each request against the vendor's published rate by default. If you negotiated your own, enter it and every figure in the console is computed from that instead.
See plansPrice your workload
Vendor rates, for referenceView all 150+ →
ModelInputOutputINFRO
Claude Sonnet 5$3.00$15$0
GPT-5.6 Terra$2.00$12$0
Gemini 3.7 Flash$0.75$3.75$0

The vendors' own published rates, per 1M tokens, billed to you by your provider. INFRO adds nothing to them. If you have negotiated your own rates, the console uses yours instead for every estimate.

Connect once. Keep provider decisions out of your product.

Change models, providers, and routing policies without rewiring your application.

One orange request pulse entering a central glass control plane and dividing into five channels that fan out to model destinations.
One request150+ models

Your application

Keep the tools you already use.

  • cURL
  • Python
  • TypeScript
  • Go
  • Dart

One API with routing and operations built in.

api.infro.io/v1

The AI ecosystem

150+ models for every core modality.

  • Text
  • Image
  • Video
  • Audio

What your application gets from one base URL.

You keep the model call. INFRO takes on the provider integrations, routing, retries, metering, cost, and controls around it.

One API, 150+ models

Run text, image, video, and audio models with one endpoint and one key. Add a model in configuration, not another SDK.

Automatic routing and failover

If a route fails before the response starts, INFRO retries on an eligible standby. Provider instability stays out of your application.

Every request traced and priced

See the model, route, latency, usage, and exact cost, attributed to a project, key, or end user.

Operational issues, surfaced

Find budgets at risk, routes without a standby, and workloads that could run on a lower-cost model.

INFRO is in private early access. The catalog, the console, and the API contract are public today, and the gateway is deployed. Keys and billing open to early customers in stages, and anything beyond v1 is marked Later. Early access details

None of this is your product.

Retries, providers, budgets, and usage tracking arrive one reasonable decision at a time. INFRO gives them one layer to live in.

Responsibility map

Without a boundary

8 jobs

One function becomes eight jobs.

Each local fix is sensible. Together, they turn product code into infrastructure code.

app/lib/ai.tslocal fixes
  1. callModel()
  2. + retries
  3. + second provider
  4. + rate limits
  5. + cost tracking
  6. + per-user attribution
  7. + budgets
  8. + routing

With INFRO

2 owners

The product stays product-shaped.

Your application keeps the decisions customers feel. The machinery around every request moves below the line.

Your product owns

PromptsModel choiceProduct logicExperience
callModel()

INFRO owns

The infrastructure around every request

CONTROL PLANE
01

Delivery

Providers · routing · retries · failover

02

Control

Rate limits · budgets · access

03

Evidence

Usage · latency · cost

Eight concerns become two clear owners.Change infrastructure below the line—not inside your product.

A route degrades. Traffic moves. Your app keeps working.

INFRO detects route failures, retries eligible requests on a healthy standby, and records every step.

A request pulse travelling the primary route, stalling where the route has gone dark, and a second pulse completing the journey on the standby route below it.
Illustrative incident summarySee the console →

0

failures reached your application

In the incident window shown in this example.

1,284

requests recovered

Failed upstream, then completed on a healthy standby.

18%

traffic moved at peak

Rerouted while the primary route recovered.

Health-aware routing and failover: · deeper anomaly detection: Later — Planned after v1. No date committed.

See exactly what happened to every request.

Three questions the console answers: what happened to a single request, whether anything broke, and where the money went.

A signal crossing five diagnostic stages in sequence, each one writing a line of measurement into a ledger plane below the path.

What happened to this request?

req_9f2c4b8e21 · claude/fable-5
Status
200
Latency
103ms
Metered
995 tok
Cost
$0.0002

Timeline

0 → 103ms

  • gateway
    103ms
  • route · primary
    14ms
  • route · standby_a
    89ms
  • stream
    62ms

Recovered. The primary route errored before any bytes reached the client, so it was retried on standby_a and completed. Your application saw a success.

Did anything break?

Route health
  • PrimaryRecovering82%
  • Standby AHealthy18%
  • Standby BReady0%
  1. 09:14Detectedp95 crossed 1.5× its baseline
  2. 09:14Actedtraffic moved to Standby A
  3. 03:40Recoveredreturning in 10% steps

Where is the money going?

One billing period

$6,142charged

$9,022at list

Saved this period$2,880−31.9%
  • Text$1,687−30%
  • Image$2,231−32%
  • Video$1,965−32%
  • Audio$259−41%

Reference prices are each vendor’s own published rate. An unverified rate is left out, never estimated.

Console views · example workload, not INFRO telemetry

Ready to put INFRO behind your own workload? Get early access →

Keep your SDK. Change three values.

If your product already uses an OpenAI client, adoption means a new API key, base URL, and model ID — not a rewrite.

  • Your existing client library
  • Request and response shapes
  • Streaming and tool calls
  • Response parsing and error handling
OpenAI-compatible text endpoint
Before — direct to the provider
import osfrom openai import OpenAI client = OpenAI(    api_key=os.environ["OPENAI_API_KEY"],) response = client.chat.completions.create(    model="gpt-5.6-terra",    messages=[{"role": "user", "content": "Summarize this release."}],)
After — through INFRO
import osfrom openai import OpenAI client = OpenAI(    api_key=os.environ["INFRO_API_KEY"],    base_url="https://api.infro.io/v1",) response = client.chat.completions.create(    model="openai/gpt-5.6-terra",    messages=[{"role": "user", "content": "Summarize this release."}],)

Three small changes: API key, base URL, and model ID. Your SDK, payloads, response parsing, streaming, tool calls, and error handling stay the same. Read the quickstart covers the media endpoints too.

Start with one key. Scale to the whole team.

Begin with a single application. Add projects, environments, budgets, policies, and roles as you grow — without migrating to a different product.

01Start

One developer, one application, one key

  • One API across 150+ models and four modalities
  • Every request visible, with its exact cost
  • A hard spend ceiling so a bug cannot run up the bill

02Grow

Several models, providers, and projects

  • Automatic failover so one provider incident does not become yours
  • Cost attributed by project, key, and end user
  • Separate keys and budgets per environment

03ScaleEnterprise — Available on an enterprise agreement rather than self-serve.

A team that needs ownership and control

  • Role-based access, model policies, and a complete audit trail
  • Single sign-on and automated provisioning
  • A contractual service-level agreement

Start is the full product, not a limited trial. Scale adds the controls and commitments larger organizations need. INFRO for Enterprise explains what an enterprise agreement adds.

Verify the claims before you trust the product

Each claim links to the catalog, documentation, product capture, or policy that supports it.

Questions, answered plainly.

Where is INFRO today — can I use it?

Not yet — though the gateway is further along than 'planned'. It is deployed at api.infro.io, the console is deployed at dash.infro.io, and the model catalog and complete API documentation are public and final enough to review. What is not open is self-serve purchasing and general key issuance: early-access accounts get keys first. Nothing on this site is purchasable today, and every capability we describe carries the release it lands in rather than being implied to exist.

Why use INFRO instead of calling OpenAI or Anthropic directly?

If you use one model from one vendor and nobody has asked what it costs, you probably should go direct — that is a real answer and we would rather give it than waste your evaluation. INFRO earns its place when at least one of these is true: you run several models across several vendors, a provider incident has taken your product down, someone has started asking which customer or feature is spending what, or provider keys have started appearing in more than one place. You still call the same providers on the same accounts; you stop owning the integration, the failover and the accounting between them.

Do I have to rewrite my application?

For text, no. The endpoint speaks the OpenAI request and response shape, so you change the base URL, the key, and the model string — two lines in most codebases — and your client library, request bodies, response parsing, and error handling are untouched. Image, video, and audio use one consistent INFRO shape across every model, which is new code, but it is new code you would otherwise have written per vendor.

And what happens if INFRO fails?

We become one more dependency — that is the honest cost of the trade, and you should price it in. Three things make it survivable. The status page runs on a separate, far simpler deployment than the gateway, so it stays up to tell you. Text speaks the OpenAI shape, so keeping a direct provider key and a base-URL switch in your config is a two-line contingency we actively recommend. And there is no published uptime SLA, because a service with no operating history cannot honestly offer one; enterprise agreements carry a contractual SLA.

How does billing work?

Two separate bills that never move together. Your providers invoice you for model usage, directly, at your own rates — INFRO is not in that transaction and takes 0% of it. INFRO invoices you for a plan: a flat monthly or annual subscription for the control plane, priced on gateway requests and organization features rather than on tokens. There is a free Developer plan with smaller allowances and no time limit, and committed volume gets an enterprise agreement with invoicing. Provider spend never appears on an INFRO invoice.

17 more answers — latency, data handling, leaving later, residency, and what happens when a provider goes down.

Give your AI product an infrastructure layer.

Integrate once. Add models, routing, observability, and spend controls as your workload grows.

Get early accessSee the console

INFRO is in private early access. See what ships in v1 and what comes later.