Giskard: pricing and features
by Giskard AI
Prices and features checked 2026-10-20 on the vendor's own pages. Anything we could not confirm says “not verified”. We don't score or rank tools: how we review.
Short answer: Giskard is best for: Python teams that want open-source agent tests and an automated vulnerability scan in one library, with a paid audit platform available later. Free tier: Open source: local Python library, bring your own LLMs, vulnerability detection across 7 risk categories, quality checks across 5 generation scenarios, community support [source, checked 2026-10-20]. Entry paid plan: Professional: price not published (contact Giskard) [source, checked 2026-10-20].
What is Giskard? Giskard's open-source library (v3, Apache-2.0) has two parts: Giskard Checks, for evals of LLM apps and agents (assertions, groundedness and LLM-as-judge checks, multi-turn scenarios), and Giskard Scan, which generates adversarial test suites from a plain-language description of your agent (prompt injection, harmful content, jailbreaks and more). The paid Professional product adds automated audits, collaboration and enterprise deployment.
What it does and how it runs
| What it tests | Checks: regressions, RAG groundedness, safety rules and multi-turn agent conversations. Scan: red-teaming across OWASP LLM Top-10 categories, prompt injection, harmful content, stereotypes and misinformation [source] |
|---|---|
| What it can test | Any sync or async callable that takes inputs and returns outputs: an LLM, a black-box agent or a multi-step pipeline [source] |
| Install | pip install giskard (checks); pip install "giskard[scan]" adds the scanner; Python 3.12+ [source] |
| Version note | v3 is a rewrite; Giskard v2 remains available but is no longer actively maintained [source] |
| Telemetry | Optional aggregated analytics; the README says no prompts or outputs are sent; opt out with DO_NOT_TRACK=1 [source] |
| Professional adds | Agent-specific automated vulnerability scans and audit reports, 50+ adversarial probes including multi-turn attacks, OWASP coverage, SSO and role-based access, scheduled alerts [source] |
| Latest release | giskard-checks v1.0.6, 9 Oct 2026 [source] |
Pricing
- Free tier
- Open source: local Python library, bring your own LLMs, vulnerability detection across 7 risk categories, quality checks across 5 generation scenarios, community support [source, checked 2026-10-20]
- Entry paid price
- Professional: price not published (contact Giskard) [source, checked 2026-10-20]
- What you pay for (meter)
- Open source is free; no price is published for Professional [source]
- Licence
- Apache-2.0 (open-source library) [source]
- Self-hosting
- The library runs locally (Python 3.12+); Professional lists on-premise, private cloud or SaaS deployment [source]
- Data retention
- not verified
Features
| OpenTelemetry | not verified |
|---|---|
| CI gating | CI/CD integration with the SDK listed for Professional [source] |
| LLM-as-judge | LLM-as-judge checks in Giskard Checks (Groundedness, Conformity, LLMJudge) [source] |
| Human annotation | Collaborative human-in-the-loop review listed for Professional [source] |
| Prompt management | not verified |
| Dataset versioning | not verified |
| SOC 2 / HIPAA | SOC2, HIPAA and GDPR compliance listed for Professional [source] |
Compare
- Promptfoo alternatives for red-teaming: Giskard next to garak and Promptfoo's own free red-teaming
- Promptfoo pricing and features · DeepEval vs Promptfoo
- LLM evaluation platforms compared (garak and Giskard are scanners and test libraries, so they are not in that 12-platform table)
Other red-teaming tools: garak