Compare vendors / AgentDojo
AgentDojo
Academic benchmark for prompt injection against AI agents
Agents and toolsOpen source
AgentDojo is a dynamic environment for testing prompt injection attacks and defenses on LLM agents, with workspace, travel, Slack and banking task suites. NIST’s CAISI called it a leading open-source framework for testing agent vulnerability and used it with attacks developed by UK AISI.
Best for: Comparing how agent defenses hold up against injected instructions.
Open source (MIT), maintained by ETH Zurich SPY Lab
- – Maps to OWASP
- – Maps to MITRE ATLAS
- – Maps to the EU AI Act
- – SOC 2 or ISO 27001
1 Legal nameAgentDojo
2 HeadquartersZurich, Switzerland
3 Founded2024
4 TeamNot disclosed
5 What they testAgents and tool use, Prompt injection and jailbreaks
6 FrameworksNot disclosed
7 CertificationsNot disclosed
8 Published pricesFree (MIT)
9 Customers and partnersNIST CAISI used it for agent hijacking evaluations and released AgentDojo-Inspect.
Worth knowing before you sign
- Latest release (0.1.35) is from Oct 2025; last commit Jun 2026.
Sources
Work at AgentDojo? Send corrections with a source to editors@bestairedteamingcompanies.com. We fix verified errors within five business days.
Compare with
- Mindgard AI red teaming firm: human-led tests plus an automated platform
- Gray Swan AI AI security lab: agent red teaming, lab evals and the Arena
- HiddenLayer AI security platform with automated red teaming and model scanning
- Trail of Bits Security research consultancy with an AI/ML assurance practice
- Pillar Security Agent security platform with agentic red teaming