Spark Chat & API

Private, Lightning-funded AI chat — talk to it in your browser or wire it into any OpenAI-compatible agent.

Lightning onlyNo signupAnonymous accounts

Quick start — chat

Open https://api.unknown.cloud/ in any browser (works great on mobile).

Tap Create account — you get an anonymous 16-digit account number.
Tap Fund and scan the Lightning QR (or paste the payment request) with any Lightning wallet. Amounts start at a few hundred sats.
Once credited, just chat. The account balance is shown in the UI.

Use the API — one command (no setup to do)

Open a terminal and run this. It configures your agent for you and verifies with a real test chat — you never type an endpoint, model, or header:

curl -fsSL https://api.unknown.cloud/docs/setup.sh | sh -s -- 0123456789012345

Replace 0123456789012345 with your account number (click the account chip in the chat to copy your exact command). The script backs up any existing agent config and only points your agent at the Spark API.

Manual details (optional)

The service exposes a standard OpenAI-compatible HTTP API. Anything that can talk to api.openai.com can be pointed here — curl, the OpenAI SDKs, Hermes Agent, and other agents.

Endpoint
https://api.unknown.cloud/v1

Model id spark-private-1
List models

Authentication

Your account number is your API key. Send it as a standard Bearer token (OpenAI style). No other header is required:

Authorization: Bearer 0123456789012345

No API key is provisioned separately — create/fund an account in the web UI to get a number, then use it here. (The legacy X-Account-Number: 0123456789012345 header also still works.)

Chat completion (curl)

Replace 0123456789012345 with your account number:

curl https://api.unknown.cloud/v1/chat/completions \
  -H "Authorization: Bearer 0123456789012345" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "spark-private-1",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Response is a standard chat.completion object with usage (prompt/completion tokens). Balance is debited per request.

Connect your Hermes agent (manual)

In your ~/.hermes/config.yaml, set the model block to a custom provider (your account number is the API key):

model:
  default: spark-private-1
  provider: custom
  base_url: https://api.unknown.cloud/v1
  api_key: "0123456789012345"

Then either restart Hermes or start a new session. Verify with:

hermes chat -q "Reply with exactly OK" -m spark-private-1

Any OpenAI-compatible client

# Python (openai SDK)
from openai import OpenAI
client = OpenAI(base_url="https://api.unknown.cloud/v1",
                api_key="0123456789012345")

Funding & limits

Topic	Detail
Funding	Pay a Lightning payment request. The only way to obtain credits is Lightning. No cards, no free accounts.
Balances	Deducted per chat completion by token usage. Check balance via the web UI or `GET /v1/balance`.
Errors	`401` missing/invalid account • `402` insufficient balance • `404` wrong path/model

Privacy

Accounts are anonymous 16-digit numbers — no email, name, or PII required. Your conversation text is sent to the model to answer; only a token-usage total is recorded per account for billing.

Beta service — behavior and pricing may change. If something behaves oddly, that's exactly the kind of report we want.