One OpenAI-compatible inference API. Cheap, drop-in.
Open models — Kimi-K2.6, Qwen3-235B, MiniMax-M2.7 — through a single endpoint. Pay by card or crypto — monthly plan or pay-as-you-go. Free key in 30 seconds.
The assistant runs on our own API — Kimi-K2.6 through Mingles Router. This chat is the product in action.
Pay by card or crypto. Plan or pay-as-you-go. 30 seconds to your first response.
https://router.mingles.ai/v1 Try the API right now
Ask our AI assistant anything — it runs on Kimi-K2.6 through this very API. No signup needed.
What you can build today
Real outcomes, not boilerplate. Pick a flow, copy a config, ship.
Vibecoding in Cursor
Code with AI without paying $20/mo. Drop your key into Cursor, pick Kimi-K2.6, write apps.
See config →No-code automations in n8n
Build agents that read your email, summarize Notion docs, post to Telegram — all visual, no code.
See config →Always-on bots with OpenClaw
Spin up a Telegram or Discord bot that thinks. No server. No DevOps. One config file.
See config →AI in your IDE with Cline
Free Claude-Code-style agent that opens files, runs commands, edits projects — pointed at one endpoint.
See config →Built for two workflows
For developers & vibecoders
One OpenAI-compatible endpoint in Cursor, Cline, n8n, OpenClaw, Hermes and custom agents. Free key, pay by card or crypto.
Start buildingFor agencies
Control LLM cost across client bots and workflows. Per-client keys, limits, usage tracking and fallback routing.
See agency backendPlug-and-play
Anything that says “OpenAI-compatible” works. One URL, one key, that's the integration.
Cheap, no lock-in
Per-token pricing well below the public-API median. Pay by card or crypto — top up only what you use.
Agents-first
Tool-calling, streaming, request tracing — everything Cursor/Cline/OpenClaw need just works.
One minute, if you do code
The official OpenAI SDK works as-is. Change one variable.
from openai import OpenAI
client = OpenAI(
base_url="https://router.mingles.ai/v1",
api_key="sk-your-free-key",
)
resp = client.chat.completions.create(
model="moonshotai/Kimi-K2.6",
messages=[{"role": "user", "content": "Make me a startup name."}],
)
print(resp.choices[0].message.content)Need Node / curl / Cursor / Cline / n8n? See all configs →
100 % open source.
Read the code, fork it, self-host it. We publish the router under a permissive license.
Frequently asked questions
▸Do I need to know how to code?
No. Most of our users run no-code agents (n8n, OpenClaw) or vibecoding tools (Cursor, Cline). The API key is a single line in a settings screen — that's the deepest you need to go.
▸What apps can I plug it into?
Anything that supports an "OpenAI-compatible API" or "custom base URL": Cursor, Cline, Continue, Aider, OpenClaw, n8n, LibreChat, Open WebUI, Jan, LM Studio, and more.
▸How is this different from ChatGPT?
ChatGPT is a chat app for one user. We are an API — the engine that powers apps and agents. If you only want to chat, use ChatGPT or our /app console. If you want apps and bots to think, you want this.
▸How do I pay?
Pay by card or crypto in the console — no KYC. Pick a monthly plan, pay-as-you-go from your balance, or start free.
▸Where can I see the code?
The router is open source: github.com/MinglesAI/gonka-proxy. Fork it, self-host it, or just read how it works.
Get your free key and start building.
30-second sign-up. Pick your tool. Paste the key. Ship.