AI Agents Engineering Track
Design, build, secure and operate real AI agents: backend foundations, the agent loop, tools and RAG, MCP and multi-agent systems, and production AgentOps — hands-on the whole way, including fixing real vulnerabilities in the DARE Vulnerable AI Suite.
This is the DARE Labs path into building production-grade AI agents — the engineering counterpart to the AI Red Teaming track. Start with AI Agents Foundations for backend basics, LLM fundamentals and structured outputs, then build your own resilient, retrying LLM client. Move into Agent Engineering to build a tool-calling agent loop from scratch, no framework, evolving it through planning, memory and retrieval. Go deeper with MCP & Multi-Agent Systems, building a real MCP server and client, then fixing an actual vulnerable MCP server in the DARE Vulnerable AI Suite from the defensive side. Finish with Production AgentOps: evaluation, testing, observability, security guardrails, deployment and cost control, including patching a real missing-authorization bug in the same shared suite. Every lab produces working code you run and verify yourself.
Track items
- 📘 AI Agents Foundations
- 🧪 Build a Resilient LLM Client Premium Requires Builder plan View plans
- 📘 Agent Engineering Premium Requires Builder plan View plans
- 🧪 Build a Tool-Calling Agent from Scratch Premium Requires Builder plan View plans
- 📘 MCP & Multi-Agent Systems Premium Requires Builder plan View plans
- 🧪 Build an MCP Server and Client Premium Requires Builder plan View plans
- 🧪 Fix a Vulnerable MCP Server Premium Requires Builder plan View plans
- 📘 Production AgentOps Premium Requires Builder plan View plans
- 🧪 Instrument and Deploy an Agent Premium Requires Builder plan View plans