Agent-first token routing

One endpoint. Any Model.

An OpenAI-compatible relay for your agents. Pass a model and routing runs on autopilot — under the budget and rules you set, no UI required. Open models, at a fraction of frontier-model cost.

Get started Read the docs

quickstart.py

from openai import OpenAI

client = OpenAI(
    base_url = "https://api.tkex.ai/v1",
    api_key  = "$TX_API_KEY",
)
resp = client.chat.completions.create(
    model    = "deepseek/deepseek-v4-pro",  # swap any model — same API
    messages = [{"role": "user", "content": "Ship it."}],
)
# → routed on demand across curated regional suppliers

Routes the open models your agents already want

DeepSeekQwenGLMOpen source+ more

Drop in

Nothing else changes.

Keep your OpenAI code. Point the base URL at us, pass a model, and you're live. Switching models is one line.

OpenAI-compatible — chat, embeddings, images, rerank.
No SDK to learn. No UI required — the API is the product.

client.py1 line changed

  from openai import OpenAI  client = OpenAI(-    base_url="https://api.openai.com/v1",+    base_url="https://api.tkex.ai/v1",     api_key="sk-***",  )

Swap one line. Everything else stays.

Any model

One line to switch.

Run your cursor down the line — every model is one line of code away. Same endpoint, same keys.

chat.py

model = "deepseek/deepseek-v4-pro" # DeepSeek V4 Pro · one line to switchmessages = [{"role": "user", "content": "Hi"}]

Autopilot

It runs on autopilot.

Pass a model — we route to the best-priced supplier, and fail over automatically.

Control

You set the budget. It holds the line.

Per-key spend limits return a structured 402 — not a surprise invoice. See every supplier in the Providers directory and blacklist any you'd skip; routing avoids it for your account.

Daily limit$25.00

Monthly limit$500.00

HTTP 402 · period_quota_exceeded

supplier Asupplier Bsupplier C

Pricing

No platform fee. Just the processor's cut.

When you add credits, you only cover Stripe's payment-processing fee — passed straight through. Comparable services add 5.5% on top.

No platform fee

At top-up you cover only Stripe's deposit fee, passed straight through — comparable services add 5.5%.

Credits land 1:1

What you add is what you can spend, 1:1 — no per-request surcharge. The Stripe fee is shown before you pay.

Routed for you, one rulebook

Pass a model and we route it across curated regional suppliers. The same pricing rules apply to every buyer in your group.

Point your first agent at it.

One endpoint, any model, your budget. No UI required.

Get started

FAQ

Questions, answered.

What does it cost to use tokenexchange?

You pay for the models you call. There's no platform fee — when you add credits you only cover Stripe's payment-processing cost, passed straight through (comparable services add 5.5%).

How do I switch models?

Change one line — the model field. Same endpoint, same OpenAI-compatible API.

Do I have to use the dashboard?

No. Everything an agent needs is in the request and the response. The console is an optional rendering of the same schema.

Can I choose which supplier serves a request?

You pass a model and we route it across curated regional suppliers on demand. See them in the Providers directory and blacklist any you'd skip — routing avoids it for your account. Which supplier handled a given call isn't shown.

What happens if I hit a spend limit?

You get a structured 402 (period_quota_exceeded) — your code handles it, not a surprise invoice. Set daily, weekly, and monthly limits per key.

Can I get a refund?

Within 24 hours of a top-up you can refund its unused credits. After 24 hours, top-ups are non-refundable; consumed credits and the payment-processing fee aren't refundable.