Classai Router API

Base URL:

Copy the default key from Dashboard → API Keys (created automatically when you sign up), buy credits or a package, then run it — this base URL and API key are all you need, no other app required:

curl https://router.classai.id/v1/chat/completions \
  -H "Authorization: Bearer sk_live_…" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-5.5",
    "messages": [
      {"role": "user", "content": "Hi! Explain AI in one sentence."}
    ]
  }'

Send the key in the Authorization: Bearer sk_live_… header. The x-api-key header is also accepted (used by the Anthropic SDK). Every account automatically gets one default key; create extra keys or fork an existing one for other apps/team members — they all share the same balance & packages, with separate usage records. Keep keys on the server or in environment variables — never in frontend code.

Send "stream": true to receive the answer incrementally via Server-Sent Events. The last chunk contains usage (token counts) — we add stream_options.include_usage automatically so usage is recorded accurately.

Prices per 1 million tokens. 1 billion kredit = $12.48. The full catalog is on the models page.

Model ID	Context	Input / 1M	Output / 1M
claude-opus-5.5Claude Opus 5.5	1M	52,8 jt kredit	264 jt kredit
claude-sonnet-5.5Claude Sonnet 5.5	1M	26,4 jt kredit	132 jt kredit
claude-opus-5Claude Opus 5	200K	66 jt kredit	330 jt kredit
claude-sonnet-5Claude Sonnet 5	200K	26,4 jt kredit	132 jt kredit
claude-sonnet-4.6Claude Sonnet 4.6	200K	39,6 jt kredit	198 jt kredit
claude-haiku-4.5Claude Haiku 4.5	200K	6,6 jt kredit	33 jt kredit
claude-haiku-5.5Claude Haiku 5.5	1M	10,8 jt kredit	33 jt kredit
glm-5.3GLM 5.3	1M	12,5 jt kredit	61,6 jt kredit
deepseek-v4.1-flashDeepSeek V4.1 Flash	1M	8,3 jt kredit	30,3 jt kredit
deepseek-v4.1-flash-1017DeepSeek V4.1 Flash 1017	1M	7,7 jt kredit	28 jt kredit
kimi-k3Kimi K3	256K	19,8 jt kredit	99 jt kredit
glm-5.2GLM 5.2	1M	12,3 jt kredit	61,6 jt kredit
minimax-m3MiniMax M3	1M	6 jt kredit	23,1 jt kredit
kimi-k2.7-codeKimi K2.7 Code	250K	8,4 jt kredit	41,8 jt kredit
minimax-m2.7MiniMax M2.7	200K	5,3 jt kredit	13,2 jt kredit
deepseek-v4-flashDeepSeek V4 Flash	1M	7,7 jt kredit	30,9 jt kredit
deepseek-v4-flash-0731DeepSeek V4 Flash 0731	1M	7,2 jt kredit	27,5 jt kredit
deepseek-v4-proDeepSeek V4 Pro	1M	20,9 jt kredit	55 jt kredit
deepseek-v4-pro-0813DeepSeek V4 Pro 0813	1M	21,4 jt kredit	55 jt kredit
glm-5.3-flashGLM 5.3 Flash	1M	7,4 jt kredit	30,8 jt kredit
glm-5.1GLM 5.1	200K	12,3 jt kredit	61,6 jt kredit
kimi-k2.6Kimi K2.6	256K	8,4 jt kredit	41,8 jt kredit
gemini-3.8-flashGemini 3.8 Flash	1M	14 jt kredit	56 jt kredit
claude-opus-4.6Claude Opus 4.6	200K	66 jt kredit	330 jt kredit
gpt-6-astraGPT-6 Astra	400K	66 jt kredit	330 jt kredit
gpt-5.6-solGPT-5.6 Sol	400K	34,2 jt kredit	110 jt kredit
gpt-6-solGPT-6 Sol	1M	34,2 jt kredit	110 jt kredit
gpt-6.1-solGPT 6.1 Sol	1.1M	34,2 jt kredit	100 jt kredit
gpt-5.6-terraGPT-5.6 Terra	1M	28,8 jt kredit	82,5 jt kredit
gemini-3.8-seriesGemini 3.8 Series	1M	9 jt kredit	43,6 jt kredit
gpt-5.6-lunaGPT-5.6 Luna	1M	12,4 jt kredit	61,9 jt kredit
qwen3.8-maxQwen 3.8 Max	1M	13,2 jt kredit	66 jt kredit
claude-opus-4.8Claude Opus 4.8	1M	66 jt kredit	330 jt kredit
coding-china-autoCoding Plan China (auto)	1M	12,5 jt kredit	61,6 jt kredit
coding-us-autoCoding Plan US (auto)	200K	30 jt kredit	150 jt kredit
gpt-6-lunaGPT-6 Luna	1M	12,4 jt kredit	61,9 jt kredit
grok-4.7Grok 4.7	500K	21,1 jt kredit	66 jt kredit
hy3Hy3	256K	9,9 jt kredit	49,5 jt kredit
qwen3.7-maxQwen 3.7 Max	1M	9,8 jt kredit	48,7 jt kredit
qwen3.7-plusQwen 3.7 Plus	1M	5,4 jt kredit	24,7 jt kredit
claude-opus-4.7Claude Opus 4.7	200K	66 jt kredit	330 jt kredit
kai-baikKai Grant · DeepSeek V4.1 Flash	1M	0 kredit	0 kredit
longcat-2.5:freeLongCat 2.5 (Free)	—	0 kredit	0 kredit
step-3.7-flash:freeStep 3.7 Flash (Free)	—	0 kredit	0 kredit
solar-mini4:freeSolar Mini 4 (Free)	—	0 kredit	0 kredit
nemotron-3-ultra:freeNemotron 3 Ultra 550B (Free)	1M	0 kredit	0 kredit
nemotron-3.5-super-vl:freeNemotron 3.5 Super VL (Free)	1M	0 kredit	0 kredit
nemotron-3.5-lightning:freeNemotron 3.5 Lightning (Free)	1M	0 kredit	0 kredit
nemotron-3-nano-omni:freeNemotron 3 Nano Omni (Free)	256K	0 kredit	0 kredit
north-mini-code:freeNorth Mini Code (Free)	256K	0 kredit	0 kredit
gemma-4-31b:freeGemma 4 31B (Free)	262K	0 kredit	0 kredit
gemma-4-26b:freeGemma 4 26B (Free)	262K	0 kredit	0 kredit
dots-3-note:freedots 3 Note (Free)	512K	0 kredit	0 kredit
flint-1.0:freeFlint 1.0 (Free)	262K	0 kredit	0 kredit
lfm-3-1.7b:freeLFM 3 1.7B (Free)	262K	0 kredit	0 kredit

Cost = input tokens × input price + output tokens × output price. Token counts come from the usage reported by the model; if a model doesn't send usage while streaming, we estimate it from the text length (≈ 4 characters per token).

Example: 1,000 input tokens + 500 output tokens on claude-opus-5.5 = $0.0023 ≈ 185 rb kredit.

Failed requests (errors from the model or the gateway) aren't charged. Some models add an internal system prompt, so input tokens can be slightly higher than the text you send.

Input tokens the model reads from cache are billed at half price — see Prompt caching.

Coding agents like Claude Code, opencode, Cline, and Cursor resend the entire conversation at every step: system instructions, the tool list, chat history, and the contents of files already read. The leading part that matches the previous request can be read by the model from cache instead of being processed again.

• Automatic, no setup. No need to instruct the AI in opencode, change your prompt, or add parameters. For Claude models, the Router adds the cache_control marker itself if your tool hasn't set it (Claude Code already does); other models cache automatically on the model side.
• Half price. Input tokens read from cache are billed at 50% of the input price. Output tokens and the uncached part are billed normally. Applies to credits and token packages bought from 10 October 2026; packages bought before that date count all tokens at full price.
• Not always 100%. How much gets cached depends on the model and the route serving it — in agent sessions usually 40–95% of input tokens. Short prompts (under ±1,000–4,000 tokens) aren't cached.

To hit the cache more often: don't change system instructions, AGENTS.md, or the tools/MCP list mid-session, and keep the session going without long pauses (the cache usually lasts a few minutes). When switching tasks, start a new session or run /compact so the conversation doesn't keep growing.

Before a prompt is sent to the model provider, the gateway replaces certain personal data with placeholders. Only the text you send is redacted (messages, system instructions, tool results) — not tool definitions, tool call arguments, images, or files.

NIKs are recognized by their pattern (province code, birth date, month), and card numbers by the card network prefix plus a Luhn check — so ordinary numbers like order IDs or timestamps aren't redacted. Redaction is deterministic (same input → same output), so prompt caching still works.

curl https://router.classai.id/v1/chat/completions \
  -H "Authorization: Bearer sk_live_…" \
  -H "x-classai-pii: strict" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-opus-5.5","messages":[{"role":"user","content":"Hello"}]}'

Redaction is a safety layer, not a guarantee: still don't send unnecessary sensitive data to any AI model.

Turn text into audio with the same key and credits. The endpoint is compatible with OpenAI /v1/audio/speech, so the OpenAI SDK works as is. A no-code version is in Voice Studio.

curl https://router.classai.id/v1/audio/speech \
  -H "Authorization: Bearer sk_live_…" \
  -H "Content-Type: application/json" \
  -d '{"model":"gemini-tts","voice":"Kore","input":"Hi, welcome to Classai!","response_format":"mp3"}' \
  --output speech.mp3

# Every active engine + price
curl https://router.classai.id/v1/audio/voices -H "Authorization: Bearer sk_live_…"

# One engine's voices (id, name, free sample audio)
curl "https://router.classai.id/v1/audio/voices?model=gemini-tts" -H "Authorization: Bearer sk_live_…"

• Billed per character (±1,000 characters ≈ 1 minute of audio), deducted from credits just like tokens. *Latest prices and voice list: GET /v1/audio/voices and GET /v1/audio/voices?model=<model>.
• Cloned voices are created in Voice Studio (a voice sample + a consent recording from the voice's owner), then used with "voice": "clone:<id>". The clone fee is charged once, the first time that voice is used on an engine.
• Dynamic pricing on our GPUs (voxcpm2, dots-tts): while a GPU is warm (busy) you pay only the per-minute price; a request that has to wake a sleeping GPU adds a one-off wake-up fee (wake_fee_rp in GET /v1/audio/voices) and its first audio takes ~2–3 minutes.
• Style & pronunciation. Add "style": "narasi" (natural, narasi, iklan, santai, berita, edukasi — natural, narration, ad, casual, news, educational). Numbers, Rupiah, times, abbreviations, and terms like AI are read with Indonesian spelling automatically, including words in your account's pronunciation dictionary; send "normalize": false to pass the text through as is. Calibrated voices automatically use their best articulation settings when voice_settings is empty.
• Formats: mp3 or wav. The response includes the x-classai-characters and x-classai-cost-micro-rp headers. Failed requests aren't charged.
• Credits only. Voice is paid from your credit balance; token packages don't cover it. An API key limited to one model only works for that model — use an unrestricted key (or one limited to the voice model).
• Errors: 400 empty or too-long text, 401 missing or wrong key, 402 not enough credits, 403 key limited to another model, 404 unknown model or cloned voice, 429 rate limit, 502/503 engine error or temporarily unavailable (not charged). Same JSON error shape as the chat endpoints.
• Listeners must be told the audio is AI-generated. Imitating someone without permission or to deceive is prohibited — your account may be closed.

Create and edit images with the same key and credits: Google Nano Banana, OpenAI GPT Image, Microsoft MAI-Image, ByteDance Seedream and FLUX. The endpoints follow OpenAI's /v1/images/generations and /v1/images/edits, so the OpenAI SDK works as is. Billed per image that comes out; failed images aren't charged. Models marked ≈ are billed by the tokens they really use (the table shows the typical price). No-code version: Image Studio.

curl https://router.classai.id/v1/images/generations \
  -H "Authorization: Bearer sk_live_…" \
  -H "Content-Type: application/json" \
  -d '{"model":"nano-banana-2","prompt":"A minimalist poster for palm-sugar iced coffee","aspect_ratio":"4:5","resolution":"2K","response_format":"url"}'

Videos run as tasks: POST /v1/videos, then poll GET /v1/videos/{id} every ~10 seconds until status is completed (1–10 minutes depending on the model), then download from url or /v1/videos/{id}/content. Models: Google Veo 3.1, ByteDance Seedance, Alibaba Wan & HappyHorse, MiniMax H3 and Oxygen. Veo has one price per video; the others are priced per second by resolution (minimum $0.416 per video). Credits are held when it starts and only charged if the video comes out. No-code version: Video Studio.

# 1) Start
curl https://router.classai.id/v1/videos -H "Authorization: Bearer sk_live_…" -H "Content-Type: application/json" \
  -d '{"model":"veo-3.1-fast","prompt":"Drone over Bali rice terraces at sunrise, nature sounds","aspect_ratio":"9:16","resolution":"720p","seconds":"8"}'
# → {"id":"…","status":"in_progress"}

# 2) Poll until completed
curl https://router.classai.id/v1/videos/<id> -H "Authorization: Bearer sk_live_…"

# 3) Download
curl https://router.classai.id/v1/videos/<id>/content -H "Authorization: Bearer sk_live_…" --output video.mp4

Status: in_progress → completed (with a url valid for 24 hours) or failed (not charged). A video running longer than 45 minutes is cancelled and its credits come back. Latest models & prices (incl. price_per_second and price_estimate): GET /v1/media/models.

OpenAI format, same key and credits: POST /v1/embeddings (vectors for search & RAG), POST /v1/rerank (order documents by relevance to a query), and POST /v1/moderations (flag harmful content). Cost = tokens × the price per 1M tokens, rounded up. Failed requests are not charged.

curl https://router.classai.id/v1/embeddings -H "Authorization: Bearer sk_live_…" -H "Content-Type: application/json" \
  -d '{"model":"text-embedding-3-small","input":["AI class price","How to join a class"]}'
# → {"data":[{"index":0,"embedding":[…]},…],"usage":{"prompt_tokens":11,"total_tokens":11}}

All API keys in your account draw from the same source. For each request, the balance of a token package for that model is used first (single-model packages take priority over bundles); if there's no package or its tokens run out, the cost is deducted from credits. No special key per package or per model is needed.

Credits work for every model and are bought in packs: Pro 250M credits $4.99 · Max 500M credits $7.49 · Ultra 1B credits $11.99. Each purchase is valid for 30 days; the credits expiring soonest are used first, and whatever is left expires when the validity period ends. Balances bought before 10 October 2026 stay intact and never expire. Each model's credit rate per token is on the credits page.

Token packages are active for 30 days. Buying the same package again before it expires adds its tokens to what's left and extends it by 30 days from the previous end date. Once the active period ends, remaining tokens expire. Packages bought before 10 October 2026 keep working until used up, but can't be extended — new packages have their own quota.

Coding Plan is used through a single auto model. You don't pick a model: Classai automatically uses whichever model is fastest at the moment, and if one is having issues, another takes over within the same request.

curl https://router.classai.id/v1/chat/completions   -H "Authorization: Bearer $CLASSAI_API_KEY"   -H "Content-Type: application/json"   -d '{"model": "coding-us-auto", "messages": [{"role": "user", "content": "Hello"}]}'

In Claude Code, opencode, Cursor, or other tools, just set the model name to coding-us-auto or coding-china-auto. Tokens are deducted from your Coding Plan quota. Other models (e.g. claude-opus-5) aren't part of Coding Plan — if called directly, they're paid from credits.

Errors follow the OpenAI format: {"error": {"message", "type", "code"}}.

Tools that support "OpenAI compatible" only need the base URL https://router.classai.id/v1. For Claude Code, use ANTHROPIC_BASE_URL without /v1 and pick a model from the catalog via ANTHROPIC_MODEL.

export ANTHROPIC_BASE_URL="https://router.classai.id"
export ANTHROPIC_AUTH_TOKEN="sk_live_…"
export ANTHROPIC_MODEL="claude-opus-5.5"
export ANTHROPIC_SMALL_FAST_MODEL="claude-opus-5.5"
claude

Using opencode? See the opencode prompt below — your API key and every model are already filled in, just copy & paste. The 9router section is only for people running their own 9router app — otherwise you can skip it.

How to set up Classai Router in popular tools — official download, install command, and setup steps one by one. Still stuck? Tap Ask Kai on the tool.

Code withkaiRecommended

Classai · CLI / terminal

Official download Ask Kai

Classai's own coding agent — terminal + Mac & Windows desktop. Sign in with Classai and go, no base URL to set.

Install · macOS / LinuxCopy

curl -fsSL https://withkai.im/install | bash

Install · Windows (PowerShell)Copy

irm https://withkai.im/install.ps1 | iex

1. 1Download the desktop app at withkai.im/download, or install the CLI with the command above.
2. 2Open the app (or type withkai in your project folder), then Sign in with Classai — Prime and Premium keys are created for you.
3. 3Start coding. Default is Classai Prime · Claude Sonnet 5.5; switch models in the model picker and top up credits right in the app.

Claude Code

Anthropic · CLI / terminal

Official download Ask Kai

Coding agent in your terminal — Anthropic format.

Install · macOS / LinuxCopy

curl -fsSL https://claude.ai/install.sh | bash

Install · Windows (PowerShell)Copy

irm https://claude.ai/install.ps1 | iex

1. 1Install Claude Code, then set these environment variables (or put them in the "env" block of ~/.claude/settings.json):
2. 2ANTHROPIC_BASE_URL = https://router.classai.id (no /v1) · ANTHROPIC_AUTH_TOKEN = your API key · ANTHROPIC_MODEL = claude-opus-5.5
3. 3Run claude and check with /status.

Terminal (macOS / Linux)Copy

export ANTHROPIC_BASE_URL="https://router.classai.id"
export ANTHROPIC_AUTH_TOKEN="API_KEY_KAMU"
export ANTHROPIC_MODEL="claude-opus-5.5"
claude

More examples in Tool integrations

OpenCode

OpenCode · CLI / terminal

Official download Ask Kai

Open-source coding agent, terminal and desktop.

Install · macOS / LinuxCopy

curl -fsSL https://opencode.ai/install | bash

Install · npmCopy

npm install -g opencode-ai

1. 1Download OpenCode (desktop) or install it from the terminal.
2. 2Open Docs → opencode prompt in the dashboard and copy the prompt (your API key and every model are filled in).
3. 3Paste it into OpenCode, then pick a Classai model with /models.

opencode prompt (key + models filled in)

Codex CLI

OpenAI · CLI / terminal

Official download Ask Kai

OpenAI's terminal coding agent — via the Responses API.

Install · npmCopy

npm install -g @openai/codex

Install · macOS / LinuxCopy

curl -fsSL https://chatgpt.com/codex/install.sh | sh

1. 1Install Codex, then create or edit ~/.codex/config.toml as below.
2. 2Put your API key in the CLASSAI_API_KEY environment variable, then run codex.

~/.codex/config.tomlCopy

model = "claude-opus-5.5"
model_provider = "classai"

[model_providers.classai]
name = "Classai Router"
base_url = "https://router.classai.id/v1"
env_key = "CLASSAI_API_KEY"
wire_api = "responses"

Hermes Agent

Nous Research · CLI / terminal

Official download Ask Kai

A personal AI agent that gets tasks done for you.

Install · macOS / LinuxCopy

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

Install · Windows (PowerShell)Copy

iex (irm https://hermes-agent.nousresearch.com/install.ps1)

1. 1Install Hermes, then in a terminal (not inside a chat) run: hermes model
2. 2Pick "Custom endpoint" and enter base URL https://router.classai.id/v1, your API key, and model claude-opus-5.5.
3. 3Switch models any time in a session with /model custom:<model-id>.

~/.hermes/config.yamlCopy

model:
  provider: custom
  base_url: https://router.classai.id/v1
  api_key: API_KEY_KAMU
  default: claude-opus-5.5

OpenClaw

OpenClaw · CLI / terminal

Official download Ask Kai

A personal AI assistant that runs on your computer and connects to your chats.

Install · macOS / LinuxCopy

curl -fsSL https://openclaw.ai/install.sh | bash

Install · Windows (PowerShell)Copy

iwr -useb https://openclaw.ai/install.ps1 | iex

1. 1Install OpenClaw, run onboarding and choose a custom API key / OpenAI-compatible.
2. 2Enter base URL https://router.classai.id/v1, your API key and model claude-opus-5.5 — or write it straight into ~/.openclaw/openclaw.json as below.

~/.openclaw/openclaw.jsonCopy

{
  models: {
    providers: {
      classai: {
        baseUrl: "https://router.classai.id/v1",
        apiKey: "API_KEY_KAMU",
        api: "openai-completions",
        models: [{ id: "claude-opus-5.5", name: "claude-opus-5.5" }],
      },
    },
  },
  agents: { defaults: { model: { primary: "classai/claude-opus-5.5" } } },
}

9router

9router · Local router

Official download Ask Kai

A local AI router on your laptop — combine Classai Router with other accounts.

Install · npmCopy

npm install -g 9router && 9router

1. 1Install and run 9router; its dashboard opens at http://localhost:20128.
2. 2Add Classai Router as an OpenAI-compatible provider — follow the 9router guide in our docs.

Full 9router guide

Cursor

Anysphere · Code editor

Official download Ask Kai

A code editor with built-in AI.

1. 1Open Cursor Settings → Models and paste your API key as the OpenAI API Key.
2. 2Turn on Override OpenAI Base URL → https://router.classai.id/v1, then add the model claude-opus-5.5.
3. 3Note: your own key works for chat only (not Tab), and Cursor's Plus plan doesn't allow your own keys.

More examples in Tool integrations

Cline

Cline · VS Code extension

Official download Ask Kai

Coding agent in VS Code (also a CLI).

Install · CLICopy

npm i -g cline

1. 1Install the Cline extension from the VS Code Marketplace.
2. 2Click the gear icon → API Provider: OpenAI Compatible.
3. 3Base URL https://router.classai.id/v1, API Key = your API key, Model ID claude-opus-5.5.

Kilo Code

Kilo · VS Code extension

Official download Ask Kai

Coding agent for VS Code and the terminal.

Install · CLICopy

npm install -g @kilocode/cli

1. 1Install Kilo Code from the VS Code Marketplace (or its CLI).
2. 2Settings (gear) → Providers tab → Custom provider, Provider API: OpenAI Compatible.
3. 3Base URL https://router.classai.id/v1, enter your API key — the model list loads by itself. Submit.

Aider

Aider · CLI / terminal

Official download Ask Kai

AI pair programming right in your terminal.

Install · macOS / LinuxCopy

curl -LsSf https://aider.chat/install.sh | sh

Install · Windows (PowerShell)Copy

powershell -ExecutionPolicy ByPass -c "irm https://aider.chat/install.ps1 | iex"

1. 1Install Aider, then set the base URL and API key in your terminal.
2. 2Run aider with openai/ in front of the model name.

Terminal (macOS / Linux)Copy

export OPENAI_API_BASE="https://router.classai.id/v1"
export OPENAI_API_KEY="API_KEY_KAMU"
aider --model openai/claude-opus-5.5

Chatbox

Chatbox AI · Chat app

Official download Ask Kai

AI chat app for desktop & phone — the easiest start.

1. 1Download Chatbox, open Settings → Model Provider → Add.
2. 2Choose OpenAI API compatible. API Host: https://router.classai.id (no /v1), leave API Path empty.
3. 3Enter your API key, add the model claude-opus-5.5, Save, then click Check.

n8n

n8n · Automation

Official download Ask Kai

Automations & chatbots with little code.

Install · npmCopy

npx n8n

1. 1Create an OpenAI credential: API Key = your API key, Base URL = https://router.classai.id/v1.
2. 2Use that credential in the OpenAI Chat Model node and set the model claude-opus-5.5.

More examples in Tool integrations

One prompt that sets up Classai Router and all its models in opencode. Copy, paste into opencode (or another AI coding agent), done — your existing config stays safe.

This prompt contains your API key, so it only shows after you sign in. Sign in to the dashboard then open Documentation → opencode prompt.

• 60 requests per minute per API key, and 180 per account (all keys and the playground together). Enterprise accounts: 1,000 per key and 3,000 per account. Every chat answer carries x-ratelimit-limit-requests / x-ratelimit-remaining-requests (per key) and the same with -account; GET /v1/account shows your limits. A 429 carries Retry-After (seconds).
• No limit on requests at once, except on the Unlimited package below.
• Input tokens are billed up to what your request holds (system, messages, tools, images) — prompts added on the way are never billed.
• Unlimited package (fair use): 60 requests per minute and at most 5 concurrent requests per account — all keys and the playground count together.
• Request body up to 25 MB; a single request can run up to 15 minutes.
• Up to 25 API keys per account (including the default key and forks); Enterprise up to 500.

Full terms on billing, packages, fair use, and refunds are in the Terms & Conditions and Privacy Policy.