Question 1

What is an AI agent scanner?

Accepted Answer

An AI agent scanner continuously monitors autonomous AI agents for security threats including compromised identities, behavioural drift, prompt injection attacks, and supply chain vulnerabilities — without requiring code instrumentation.

Question 2

What threats does Quin detect?

Accepted Answer

Quin detects prompt injection and jailbreak attempts, behavioural anomalies compared to baseline, over-privileged or misconfigured agent identities, unregistered agents, and agents with dependencies on models known to have vulnerabilities.

Question 3

Does Quin require instrumentation?

Accepted Answer

No. Quin scans your agent ecosystem from the outside without requiring code changes to your agents or pipelines.

Question 4

Is Quin compatible with LangChain, CrewAI, or other frameworks?

Accepted Answer

Yes. Quin is framework-agnostic and works with any agent runtime including LangChain, CrewAI, AutoGen, and custom implementations.

Question 5

Does my source code or system prompts get sent to Gaincontrol?

Accepted Answer

No. Quin runs entirely in your environment using your own LLM API key (Anthropic, OpenAI, Google, or a local Ollama model). Code and extracted prompts go directly from your machine to your chosen LLM provider for classification and synthesis — never to Gaincontrol. Run with --no-llm for a fully static scan that makes zero external LLM calls; the only network call in that mode is an unauthenticated lookup to OSV.dev for known CVEs.

Question 6

How accurate is Quin, and can it produce false positives or negatives?

Accepted Answer

Quin is an early public release (v0.1.0b2) under active development. Like any static/LLM-assisted analysis tool, it can produce false positives and false negatives, and results should be independently reviewed before being used for security, compliance, or architectural decisions. Every finding in a report includes a confidence score so you can triage accordingly.

Question 7

How long does a scan take, and how is Quin versioned?

Accepted Answer

Quin runs its 13 scanner plugins in parallel, then two LLM passes (classification, then synthesis) — scan time scales with repo size and your chosen LLM provider's latency rather than a fixed duration. Quin follows Semantic Versioning; the current release is a beta (0.1.0b2), with every change tracked in the project's changelog.

Nothing in your codebase
is hidden from Quin.

Complete AI agent inventory.
Zero manual effort.

AI indicator detection

System prompt discovery

LLM-powered intent analysis

Structured compliance reports

Free forever.
Self-hosted, no limits.

GitHub org scanning.
SIEM integration. SSO.

Frequently asked questions

Nothing in your codebaseis hidden from Quin.

Complete AI agent inventory.Zero manual effort.