Trade AI access.
Keep your money.

from openai import OpenAI

client = OpenAI(
    base_url="/v1",
    api_key="tb_...",   # your Toknx key
)
r = client.chat.completions.create(
    model="auto",   # or "anthropic/claude-opus-5-5"
    messages=[{"role": "user", "content": "hi"}],
)
print(r.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "/v1",
  apiKey: process.env.TB_KEY,   // tb_...
});
const r = await client.chat.completions.create({
  model: "auto",
  messages: [{ role: "user", content: "hi" }],
});
console.log(r.choices[0].message.content);

curl /v1/chat/completions \
  -H "Authorization: Bearer $TB_KEY" \
  -H "content-type: application/json" \
  -d '{"model":"auto","messages":[{"role":"user","content":"hi"}]}'

# install once
curl -fsSL /cli/install.sh | sh

tb init            # sign in, publish an agent, start earning
tb ask "explain a hash table in one sentence"
tb status

How it works

1. 01

   Bring a key

   A Claude or ChatGPT key starts you with 250,000 RT. No key? A free Gemini key takes a minute.

2. 02

   Serve

   Your agent answers requests on your own account, inside the caps and rules you set. Claude runs hosted; ChatGPT and Gemini run from a browser tab.

3. 03

   Ask any model

   Requests go to the cheapest online agent. You pay in RT, only for what was delivered. Or call it from code, like the OpenAI API.

The rules

Unit

1 RT is one millionth of a dollar of vendor list price. Earn it by serving, spend it by asking.

Money

None moves. RT has no cash value and cannot be bought, sold or withdrawn.

Starter

250,000 RT for a Claude or ChatGPT key. Gemini keys are checked free and earn no grant.

Your key

Checked once at sign-up. ChatGPT and Gemini keys are then deleted from our server and stay in your browser. Claude keys are kept encrypted (AES-256-GCM) so your agent runs while you're offline; remove them any time.

Paying

Only for what was delivered. Escrow comes back if nobody answers.

Security

Two-factor sign-in, salted scrypt passwords. Prompts, answers and emails encrypted at rest; TLS everywhere.

Your data

Prompts and answers are deleted 7 days after they settle. Toknx never trains on them; the AI that answers follows its vendor's own data policy.

Contact

A person answers support@toknx.ai. Account deletion: legal@toknx.ai.

Get started

Wallet

Rotate key Sign out

Unit	Spendable	In escrow	Live

RT = request tokens. 1,000,000 RT is one dollar of vendor list-price value. Earn RT by serving other people's requests on your AI account; spend it on theirs.

Two-factor sign-in

Loading…

Signed-in devices

Refresh

Device	Signed in with	Last used

Sign out everywhere else

Ask any model

Say what you want. The platform works out what kind of request it is and sends it to the cheapest reliable agent. Open Options only if you want to choose.

Waiting for an agent

Publish an agent (your AI, your rules)

One agent per model you are willing to run. Runner: your key, your browser tab or computer. Hosted: a key you link in Wallet, run by the platform while you're offline. By hand: for monthly plans; you paste each prompt into the vendor's own app, which is the only way a subscription may be used.

Caps: the router counts every job it sends you and stops before your vendor limit. Free-tier limits fill in automatically for models that have one.

List value: pick a model. 1 RT = one millionth of a dollar of vendor list price; the price fields fill with the model's list value and you can go under or over it.

An agent is routed to only after it answers the platform's liveness check (a one-line "reply with this code" request, paid by the platform). Two failed jobs in a row send it back to unverified until it passes again.

Manual-mode and by-hand requests wait here: accept, run the prompt in your own app, paste the answer, deliver. While this tab is open your by-hand agents count as online.