Start from the problem, not the product

Senthex is an AI firewall and verifiable audit trailin front of your LLMs and agents. Most buyers don't arrive searching for a “LLM firewall” — they arrive with a problem to solve. Here are the five we hear most, each answered with what actually ships today and what is the v1.1.7 pilot.


Frequently asked questions

What is Senthex?

Senthex is an AI firewall — a reverse proxy you reach by changing one base_url — plus a verifiable audit trail, placed in front of your LLMs and agents. It detects prompt injection, redacts PII, and records what your AI did. It is EU-hosted and self-hostable.

Is Senthex a firewall or an audit tool?

Both. It runs 26 best-effort input/output shields on every call (the firewall) and keeps a metadata-first record of each call (the audit trail). In the v1.1.7 pilot, that record is chained with SHA-256 and anchored to an RFC 3161 timestamp so it can be verified offline.

What is shipping today versus the pilot?

Shipping: the proxy, the shields, metadata logging, EU hosting and self-host. Pilot (v1.1.7): the verifiable, tamper-evident chain — SHA-256 + RFC 3161 timestamp + offline verifier. We label this clearly everywhere so nothing surprises you in a technical review.

Which EU AI Act articles does Senthex help with?

Record-keeping maps to Articles 12 and 19 (generate and retain logs, at least six months). The shields are a technical control that maps to Article 15(5) cybersecurity. No single vendor makes you compliant with Article 15 — that is a system-level duty.

How do I start?

Change your base_url to Senthex and add an X-Senthex-Key header — your existing SDK and API key are unchanged. The Free plan is 1,000 requests/month with no credit card. Detection is best-effort, not exhaustive (OWASP ranks prompt injection LLM01).