01 / 05
Own Compute Clusters for Open-Weight Models
Selected open-weight models run on TokGate-managed compute under clear commercial terms. We control the infrastructure directly instead of depending on anonymous rented capacity.
Connect global models,deliver real business value
Built on years of cross-border business experience, we provide standardized access to global AI models through one API, helping teams integrate faster and operate with confidence.
Model Playground
Explore what this model can do.
Try GPT-6 Astra
- 70+
- Models available
- 99.9%
- Target availability
- <200ms
- Typical gateway latency
- 0%
- Request body retention
Access leading AI models through a single TokGate API, with support for the agents you already use.
MODEL PROVIDERS Unified through one API
Google
Alibaba
Z.ai
DeepSeek
SpaceXAI
Moonshot
OpenAI
Anthropic
ByteDance
MiniMax
Xiaomi
POPULAR AGENTS Works with TokGate API
OpenClaw
Hermes Agent
Cline
OpenHands
Kilo Code
Continue
Cherry Studio
BYOK CLI TOOLS Custom API key & endpoint
PiClaude Code
OpenCode
Aider
Qwen Code
Crush
goose
Unified, visible, dependable
Why TokGate?
02 / 05
Commercial Open Licenses, Clearly Verified
Every self-hosted open model is reviewed against its commercial license before deployment. The supply chain stays explicit, traceable and suitable for real business use.
03 / 05
Multi-Region Edge Nodes with Automatic Failover
Six regional entry points and multiple upstream routes keep requests moving. When one source becomes unavailable, traffic can move to a healthy route toward our 99.9% availability target.
04 / 05
Settlement Documents Matched to Your Situation
Available contracts, invoices and settlement materials vary by contracting entity, payment method and region. Confirm the exact document set with our business team before purchase.
05 / 05
Request Bodies Are Not Stored by Default
TokGate retains only the usage metadata required for billing and operations, not your conversation bodies. Upstream providers process requests under their own published data-handling policies.
Built for production
Production-ready multi-model integration
Three protocols. One gateway.
Keep the OpenAI Chat Completions, Anthropic Messages or native Gemini format—swap the endpoint and key, with streaming and tool calls supported.
70+ models. One catalog.
Access Kimi, GLM, DeepSeek, Qwen, MiniMax, GPT, Claude and Gemini, with published specifications and context sizes.
Transparent Billing. Published Rates.
Top up at 1 USD = 200 Credits, with Credits deducted for each call at the model’s published rate and top-ups, usage and balance visible in real time.
Regional routing, built for continuity.
Edge routing and multi-source failover target 99.9% availability, with typical gateway latency under 200ms.
JYK Ecommerce Limited (Hong Kong) · Overseas settlement · Results vary by region and model.Top up Credits
Choose an amount
Top up at 1 USD = 200 Credits, with Credits deducted for each call.
FAQ / SUPPORT
What you might want to know before using tokgate.io
No. tokgate.io is a unified model API gateway that brings selected self-hosted open-source models and compliantly sourced upstream model channels into one system of authentication, routing and metering. The actual supply method of each model is as labeled in the model catalog.
One key away
Start Building with TokGate Today
Join developers using AI through one simple, transparent and production-ready gateway.
// JDK 21 HttpClient — hit the OpenAI-compatible endpoint, no SDK required
var json = """
{"model": "kimi-k2.5", "messages": [{"role": "user", "content": "Hello"}]}""";
var req = HttpRequest.newBuilder()
.uri(URI.create("https://api.tokgate.io/v1/chat/completions"))
.header("Authorization", "Bearer " + System.getenv("TOKGATE_API_KEY"))
.header("Content-Type", "application/json")
.POST(HttpRequest.BodyPublishers.ofString(json))
.build();
var resp = HttpClient.newHttpClient()
.send(req, HttpResponse.BodyHandlers.ofString());
System.out.println(resp.body());import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.tokgate.io/v1",
apiKey: process.env.TOKGATE_API_KEY,
});
// Stream the reply back — lower time to first token
const stream = await client.chat.completions.create({
model: "deepseek-v4-pro",
messages: [{ role: "user", content: "Hello" }],
stream: true,
});
for await (const c of stream) process.stdout.write(c.choices[0]?.delta?.content ?? "");# pip install openai — moving over from OpenAI is a one-line base_url change
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.tokgate.io/v1",
api_key=os.environ["TOKGATE_API_KEY"],
)
resp = client.chat.completions.create(
model="kimi-k2.5",
messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)# Any HTTP client will do — no SDK, no wrapper
curl https://api.tokgate.io/v1/chat/completions \
-H "Authorization: Bearer $TOKGATE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "glm-5.1",
"messages": [{"role": "user", "content": "Hello"}],
"temperature": 0.7
}' | jq -r '.choices[0].message.content'
# Swap models by editing one field; Credits are billed at each model's own rate