Koreshield
Screens customer text and retrieved data for prompt injection before your AI agent acts
Koreshield checks three places an AI workflow can be hijacked: customer input, retrieved documents like tickets or RAG results, and the tool calls an agent is about to make. It starts in detect mode next to live traffic — logging threats without blocking anything — so you can review false positives before switching a boundary to enforce. The team lists what it doesn't do: replace your authorization system, promise complete attack coverage, inspect files or images yet, or make high-risk agents autonomous.
What holds up
- +Publishes an honest limits list — it doesn't claim complete attack coverage
- +Detect mode runs beside live traffic first, so you see false positives before it blocks
- +Runs as a self-hosted FastAPI + PostgreSQL service, not a black-box SaaS call
Mind the limits
- −Doesn't inspect arbitrary files or images yet — only text-based inputs are covered today
- −No pricing on the page — only a free evaluation and a demo request
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/koreshield/"><img src="https://stillworks.dev/badge/koreshield.svg" alt="Picked by StillWorks" width="230" height="54"></a>