Skip to content

AIGoat v2.0: From LLM Chatbot to Agents and MCP

Latest
Latest

Choose a tag to compare

View all tags
@iamfarooqh iamfarooqh released this 10 Oct 05:38
Immutable release. Only release title and notes can be modified.
v2.0
e1c538b
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

The largest AIGoat release so far. v2.0 adds the attack surfaces that appear once a model can act: a tool-calling agent, an MCP host and client with real MCP servers, agent memory, and an end-to-end kill chain.

Highlights

  • 46 labs (23 LLM, 12 Agentic, 11 MCP) across three OWASP frameworks
  • Agentic Kill Chain capstone with Vulnerable, Defended and Guardrailed modes
  • Learn pages, Threat Modeling workbench, unified hub UI
  • Google Colab notebook and a Docker setup reachable on your network
  • Alembic migrations and a safer start.sh

Get started

./scripts/start.sh (Python 3.11, Node 18+, Ollama)
Tool-calling labs need a model with native tool calls: ollama pull qwen3.5:9b

Upgrade notes

Lab IDs follow the 2026 numbering, so update old bookmarks and handouts. Existing kill chain users should press Hard reset once.

Full notes: RELEASE_NOTES_v2.0.md
Changes since v0.3: v0.3...v2.0

Assets 5
Loading