Giskard

Giskard, made by Giskard AI (Europe), is an automated testing platform that continuously scans LLM agents for both security vulnerabilities (prompt injection, data disclosure) and quality issues (hallucinations, contradictions), before and after deployment. Who it's for: teams deploying LLM agents who need continuous, automated red teaming rather than a one-time pre-launch security review, especially in regulated industries needing GDPR/SOC 2/HIPAA compliance support. Strengths: black-box testing (working via API endpoints without needing internal agent details) means it can test agents regardless of how they're built internally, and 50+ adversarial probes with severity ranking give a real, prioritized picture of what to fix first rather than an undifferentiated list. Weaknesses and limitations: the free Solo tier is open source on GitHub; the Hub (Enterprise) tier's real pricing isn't published on the main page. Real-world scenario: a team running an LLM agent in production wants ongoing, automated testing for both security vulnerabilities and quality regressions as the underlying model or prompts change over time.