Contribute your data

TKX Documentation

TKX aggregates the AI router & compute market by measurement, not by claims. The primary way onto the boards is to let TKX read your data: expose your prices on a standard endpoint and our collector ingests and verifies you automatically, every hour. You can also consume the aggregated data through our free public API.

✓

Run a router or aggregator? Get your data in. Expose your prices on a standard /v1/models endpoint — TKX lists and measures you automatically, refreshed hourly, no adapter and no fee. It is the fastest way onto the boards, and how every aggregator contributes. → Implement the Open Router Standard · Get listed

Contribute your data (recommended)

You expose, we measure. Publish public pricing on your /v1/models in any accepted dialect and TKX pulls it hourly — no work on our side, no adapter on yours. This is how novita.ai and Requesty are already on the boards. Two steps:

1/docs/standard/Implement the Open Router Standard — expose pricing fields.

2/docs/register/Register your endpoint — we validate and start ingesting.

Who can contribute — by role

You are	Data you contribute	How
Router / Aggregator	Per-model prices (any of 4 dialects) + optional 24h usage	Public `/v1/models` (or one-api `/api/pricing`) → register; usage via a public `tkx-usage.json` — see Usage reporting
Inference provider	Your own serving prices (and measured TTFT/TPS via our probes)	Same standard, same form — reviewed as kind=provider
Model lab / vendor	Official list prices for your models (evidence tier: official)	Public pricing endpoint (dialect A/B) → register and note you are the model owner; we cross-check against open pricebooks
GPU cloud	Spot / on-demand GPU rates for the Compute board	Public price API → email echo@minax.io; live examples: Vast.ai, RunPod
App / agent builder	Token-demand attribution on the Apps board	Attribute your traffic in the OpenRouter / Vercel AI Gateway ecosystems (public leaderboards we ingest) — no direct feed needed

Every contributed number carries its evidence tier (official / API-measured / probe / self-reported) and passes sanity valves before it can appear. Self-reported data is always labeled and never mixed into measured figures or indices.

What TKX gives you

Provider & aggregator trust boards — coverage, cheapest-wins, measured markup vs list price, uptime/latency where probed.
Model rankings — per-model cheapest offer across every provider, with 30-day price history.
Price indices — TTPI (token price index) and TKX-GI (GPU spot), with reproducible methodology.
Market events — 24h price-change stream, top movers, trending apps.

The core principle

Public list prices are a commodity — a dozen sites copy them. What is scarce is verified data: what a provider actually charges, actually delivers, actually keeps online. So every data point carries an evidence tag, and self-reported numbers are always down-ranked.

“They expose, we measure. We never accept self-reported numbers.” — this is why the Open Router Standard asks routers to expose pricing rather than push it.

Consume TKX data

The aggregated data is a free public API — no key, no signup, static JSON under https://tkx.org/api/v1/, CDN-served and CORS-enabled (attribution required).

GET/docs/quickstart/Your first request in 30 seconds.

GET/docs/api/Full API reference with live examples.

The Open Router Standard