Code review, analysis, and security research

Fifty times cheaper on the tokens your agent actually burns

Agents spend most of their tokens on input: long context, tool results, the same file read twice. Top tier models charge $5.00 per million input tokens. Mirai 1 charges $0.10. That gap is Mirai 1, not the flash sampler. Six flash models sit beside it, billed at list cost on the API, not at a fake discount. The endpoint speaks the chat completions contract, so point Hermes, OpenClaw, or whichever harness you already run at satuapps.com/v1.

Get an API key

Sign in to get a key, or start on a trial key that runs for 24 hours.

Drops into the harness you already run

The endpoint follows the chat completions contract, so Hermes, OpenClaw, and anything else that speaks it work as they are. Swap base_url and your API key, then keep your existing integration.

API

Chat completions compatible API at satuapps.com/v1, with tool calling and per token billing. Point an existing client at a new base URL and keep the rest of your integration.

API docs →

API key

Sign in and the key appears on this page. In a hurry? A trial key runs straight away, good for 24 hours with a token ceiling.

Get an API key →

Priced for loops, not one question

Long loops, repeated reads, tool calls and retries. The endpoint is shaped for that work, not for one question and one answer.

Tool calling passes through

Send tools and tool_choice exactly as your harness already builds them. Mirai answers with the calls; your loop keeps running the tools.

250 thousand tokens of context

A long agent conversation carrying file contents and tool output fits without trimming history or rewriting your prompt strategy.

Streaming by default

Replies arrive as server sent events, so a long answer shows up while it is written instead of after it finishes.

The loop is what costs money

Every retry, every tool result, every extra read of the same file is billed at $0.10 per million input tokens. That is what makes leaving an agent running affordable.

Keys, sessions, and measurement

An API key is a standing charge on your account, so it is treated that way. Access is verified, sessions are protected, and every request is measured.

Encryption in transit

All traffic to Mirai uses HTTPS. Prompts and responses are encrypted between your browser and our infrastructure.

How you sign in

Sign in with a one time code, or with an email and a password. A password is never kept the way you typed it: what we store is a scrypt hash, which cannot be turned back into it.

Accountability and scope

Focused profiles require reviewed authorization. Every request is measured for billing and abuse prevention.

Pricing

The six flash models are billed at list cost, so the API price is the catalogue price. Mirai 1 costs $0.10 input and $0.80 output per million tokens. Mirai 2 stays paid. Add credit from $5 and the balance does not expire.

How billing works

The six flash models are billed at list cost, with no invented discount on top. Mirai 1 uses $0.10 input and $0.80 output per million tokens. Mirai 2 stays paid. Add credit from $5. The balance does not expire.

Add credit

Mirai 1

mirai-1

Everyday conversation, drafting, summarizing, and analysis. 200 web chats a day, bill covered. API: 50 free requests a day per key, then $0.10 input and $0.80 output per million tokens.

Input per 1M tokens
$0.10
Output per 1M tokens
$0.80

Mirai 2

mirai-2

Deeper reasoning and complex coding. Needs an active plan or balance to use.

Input per 1M tokens
$0.50
Output per 1M tokens
$1.50

Flash models to try

Six named models, billed at list cost on the API, with no invented discount.

Google Gemini 3.8 Flash

gemini-3.8-flash

Flash model for everyday web chat. 50 web chats a day. API billed at list cost.

Input per 1M tokens
$0.75
Output per 1M tokens
$3.75

Plans

Monthly plans

Fixed monthly pricing, with usage measured from the context and output each request actually uses.

Starter

$10 per month

$10 a month for 12.5 million tokens. Fits writing, short research, and lighter daily agent use.

See Starter
Most popular

Plus

$20 per month

$20 a month for 25 million tokens. Twice the Starter budget, same rolling 4 hour refill.

See Plus

Pro

$99 per month

$99 a month for 125 million tokens. Ten times Starter, for long agent loops that stay up.

See Pro

Every plan includes

  • Usage returns on a rolling 4 hour basis
  • Separate weekly and monthly safeguards
  • Mirai 1 and six flash models on the API
  • Optional credit for additional usage

Usage follows the work, not a message count

A short question uses less than a long document or a detailed code review. Your allowance is tracked from actual token usage and becomes available again on a rolling 4 hour basis, with separate weekly and monthly safeguards.

Call it from your own code

Mirai follows the chat completions contract. Point your harness at a new base URL, add your key, and keep the rest of your integration familiar. The public models are Mirai 1, Mirai 2, and six flash models.

curl https://satuapps.com/v1/chat/completions \
  -H "Authorization: Bearer $MIRAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mirai-1",
    "messages": [{"role": "user", "content": "Halo!"}]
  }'
Read the full documentation →

What changes, and what does not

  • Chat completions compatible: change base_url, keep your harness and tools.
  • Transparent pricing from the public catalogue, billed per million tokens.
  • Balance topped up from $5 that does not expire, with no monthly minimum.
  • Works with common agent harnesses such as Hermes and OpenClaw.
  • 250 thousand token context on Mirai 1, 500 thousand on Mirai 2, with tool calling passed through, so long agent loops keep their state.

Reviewed access for authorized security work

For security audits, bug bounty, red team exercises, and defensive research. Submit your organization, scope, and intended use for review. Approved accounts receive a focused profile tuned for authorized security work.

Request focused access

Used in the open

  • satuapps builds Mirai and runs its own products on the same API.

  • Every price shown here comes from the same catalogue the API bills from.

  • Four languages: English, Indonesian, Japanese, and Spanish, with light and dark themes.

Point your agent at Mirai

Copy an API key and plug it into the harness you already use.

Get your API key