Indonesian-Language AI Chatbot with Guardrails.
One OpenAI-compatible endpoint for hundreds of LLMs — with NIK, email, and credit-card masking before data reaches the LLM, Indonesian-language content moderation, and data stored in Jakarta.
Why chatbots need a gateway
Customer-service chatbots receive sensitive data: NIK, email addresses, credit card numbers, and — when enabled — phone numbers, typed freely by users. If those messages are forwarded raw to an overseas LLM API, your users' personal data goes with them. Anoman sits between your app and the LLM provider: every request passes through a guardrails pipeline (prompt injection detection, PII masking, content moderation) before a single token leaves your boundary.
Automatic NIK, email & credit-card masking
Our PII scanner detects NIK (KTP), email, and credit cards in every request — then redacts, tokenizes, or blocks them before they reach the LLM provider (phone, IP, and URL detection available as opt-in).
Indonesian-language content moderation
The content moderation filter supports both Indonesian and English — a combination of keywords and ML classification — so your chatbot never forwards harmful content in either language.
Prompt injection detection
Every user message is checked by an ML classifier before it is processed. Jailbreak attempts or hidden instructions are blocked with HTTP 403 — not forwarded to the model.
Stored in Jakarta
Traces and usage logs are stored in a Jakarta data center; prompt and response content is stored only if you turn on content capture (off by default).
Without changing code
The endpoint is OpenAI-compatible — just change the base URL to https://api.anoman.io/v1 and your chatbot is instantly protected. The OpenAI SDK, LangChain, or any agent framework keeps working as usual.
curl https://api.anoman.io/v1/chat/completions \
-H "Authorization: Bearer anm-sk-..." \
-d '{
"model": "gpt-4o-mini",
"messages": [{
"role": "user",
"content": "Please check my order status. NIK 3201234567890001, email [email protected]"
}]
}'# PII is masked BEFORE the request is forwarded to the LLM provider
Please check my order status. NIK <ID_NIK>, email <EMAIL_ADDRESS>
# Response headers confirm what was masked:
X-Anoman-Guardrail-Pii: masked 2 entities (1 ID_NIK, 1 EMAIL_ADDRESS)Four PII handling modes
- Redact — replace with a placeholder such as <ID_NIK>
- Tokenize — reversible tokens; the original value is restored in the final response
- Synthetic — realistic fake values so the model keeps full context
- Block — reject the request when a sensitive entity is present
Data stored in Jakarta
Traces and usage logs are stored in a Jakarta data center; prompt and response content is stored only if you turn on content capture (off by default).
Start free with a daily guarded allowance plus a premium-model trial credit, or see the tier pricing comparison. Related: UU PDP readiness for AI apps and cut LLM costs.
Frequently asked questions
Do I need to change my chatbot code to use Anoman?
No. Anoman's endpoint is OpenAI-compatible — just point the base URL to https://api.anoman.io/v1 and your chatbot is instantly protected. The OpenAI SDK, LangChain, or any agent framework keeps working as usual, with access to hundreds of LLMs through a single endpoint.
How does Anoman protect chatbot users' personal data?
Our PII scanner detects NIK (KTP), email, and credit cards in every request, then redacts, tokenizes, replaces them with synthetic values, or blocks them before the data reaches the LLM provider. Phone, IP, and URL detection are available as opt-in. The X-Anoman-Guardrail-Pii response header confirms which entities were masked.
Does the content moderation support Indonesian?
Yes. Anoman's content moderation filter works in both Indonesian and English — a combination of keywords and ML classification. In the same pipeline, an ML classifier checks every message for prompt injection and blocks jailbreak attempts with HTTP 403.