Inference provided

Texas-first AI inference. American-operated.

FunkyRouter is building OpenAI-compatible access to powerful, economical open-weight AI models hosted exclusively in Texas for Texas users. Join the waitlist for private pilot onboarding.

Join pilot waitlist Preview the API

Private pilot waitlist · Shared capacity · Final limits pending measurement

Your request in · model response out

Your appOpenAI SDK

FunkyRouterGateway target

Texas inference target

Open-weight modelsApple silicon

Company controlledPlanned Texas fleet

Launch path in verification

01 — What you get

A direct route to useful local infrastructure.

One focused model, a familiar developer interface, and a beta sized around real Texas customers—not an anonymous compute marketplace.

Interface

Keep the OpenAI SDK

Change the base URL and model ID. Streaming and non-streaming Chat Completions are the first contract we are shipping.

Model catalog

Measured routes

Start with a focused catalog of powerful, economical open-weight models and limits measured on the exact deployed hardware.

Operator

A team you can reach

YEETUM LLC operates the FunkyRouter public pilot, with business-hours support planned for the first customers.

Capacity

Truthful beta limits

Shared capacity, explicit token and concurrency limits, early overload responses, and no high-availability promise from one node.

02 — Why Texas

Built close to the people it serves.

FunkyRouter starts with company-controlled Apple silicon and a direct Texas inference path. That keeps the first product small, inspectable, and accountable.

The marketing site can live globally at the edge. Customer inference will not be described as Texas-hosted until the public API gateway, broker, metadata store, backups, and model execution path are physically verified in Texas.

Texas first

Local hardware. American operation. Clear limits.

Home base: Fort Worth, Texas
Launch fleet: Company-controlled Macs
Availability: Public pilot, no HA SLA

03 — Data handling

Zero content retention is the launch target.

ZDR does not mean no records exist. The launch policy separates transient inference content from the metadata needed to meter, secure, and support the service.

Inference content

The launch policy is zero persistent retention for prompts and completions. Content is processed transiently to answer the request.

Usage metadata

Request IDs, model revision, token counts, timing, status, and billing records remain so the service can operate and reconcile usage.

Training

FunkyRouter does not plan to use customer prompts or completions to train, fine-tune, evaluate, or improve models.

Verification gate: These are launch controls under implementation. Public ZDR claims become final only after storage, log, trace, crash, and backup canary scans pass.

04 — Developer experience

OpenAI-compatible by design.

Keep the SDK and streaming loop you already use. The exact quickstart below is the public pilot contract; capacity and model availability remain explicitly limited.

python · contract preview

from openai import OpenAI

client = OpenAI(
    base_url="https://api.funkyrouter.ai/v1",
    api_key="your-api-key",
)

response = client.chat.completions.create(
    model="your-model-id",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True,
)

for chunk in response:
    content = chunk.choices[0].delta.content
    if content:
        print(content, end="")

Streaming SSE in the OpenAI format

Models Open-weight pilot catalog

Status Private pilot waitlist open

05 — Provide a Mac

Link Apple silicon to your workspace.

Existing pilot users can install the Swift CLI without sudo, link a Mac, and approve its device code from the FunkyRouter app.

terminal · Apple silicon macOS

curl -fsSL --proto '=https' --tlsv1.2 https://funkyrouter.ai/install.sh | sh

After installation

Run ~/.local/bin/funkyrouter login and open the approval URL it prints.
Approve the code in your active FunkyRouter workspace.
Run ~/.local/bin/funkyrouter status to confirm the Mac is registered.

Pilot limitation: This arm64 build has an ad-hoc signature and is not notarized. It does not include the production MLX runtime. Registration is non-serving and does not enable request routing or earnings.

06 — Texas pilot

Start with one customer, measured models, and honest capacity.

The first commercial offer is designed for a Texas team that wants a reachable operator and predictable access—not a vague promise of unlimited infrastructure.

Pilot offer

$499 / month

FunkyRouter Texas Pilot

Open-weight AI models through an OpenAI-compatible API
Defined monthly token and concurrency limits
Shared capacity with early overload responses
Business-hours support planned from Texas
No high-availability SLA until a second node is proven

Join pilot waitlist

Final allowance, pricing, and availability follow measured capacity on the deployed artifact.

07 — Pilot updates

Follow measured capacity and availability.

Join the waitlist for capacity news and pilot invitations. New account signup is paused. Existing users can sign in.