explainx / curriculum sample

AI agents & MCP curriculum — tools, safety, and delivery

For platform teams and product orgs rolling out agents beyond a PoC. Emphasizes evals, tool boundaries, and failure modes—matching how we teach agent skills on explainx.ai.

About the Instructor

Yash Thakker

AI Instructor & Product Leader

Yash Thakker has 12+ years of experience building AI products and has taught 160,000+ students across 50+ courses. He facilitates corporate AI training for enterprises including Tata, PayPal, and Fortune 500 teams. Yash holds an MBA from SIMSREE and a B.Tech in Information Technology. Based in Mumbai, he delivers programs globally, specializing in Claude AI, generative AI, and practical AI implementation for regulated industries.

Credentials

  • MBA, SIMSREE (Sydenham Institute of Management Studies)
  • B.Tech, Information Technology, University of Mumbai
  • 12+ years building AI products
  • 160,000+ students trained across 50+ courses

program objectives

  • Separate demo stories from production agents with explicit tool scopes and timeouts.
  • Design MCP-style integrations with least-privilege patterns and observability hooks.
  • Stand up a minimal evaluation suite (tasks, graders, regression sets) before expanding features.
  • Connect engineers and PMs to shared vocabulary so leadership reviews stay technical, not mystical.

how we deliver

  1. 1

    Discovery call & problem framing

    We align on sponsors, success metrics, and constraints (2026 tool landscape, data rules, procurement gates) before anything is scheduled company-wide.

  2. 2

    Stakeholder interviews & day-in-the-life context

    Short conversations with practitioners (not only leadership) so scenarios reflect real workflows—not generic slide demos.

  3. 3

    Curriculum design & artifacts

    Modular agenda, exercise scripts, evaluation rubrics, and governance checkpoints matched to your vocabulary (banking, FMCG, engineering, etc.).

  4. 4

    Engaged, hands-on delivery

    Facilitation-led sessions with live exercises, breakout prompts, and documented failure modes—minimum passive lecture time.

  5. 5

    Post-session support: documentation & next steps

    Written recap, pilot backlog, links to explainx.ai courses for scaled upskilling, and optional office hours so momentum doesn’t stop at the workshop.

modules

Agent anatomy — planner, tools, memory, escalation

Avoid science-fair wiring diagrams in critical paths.

session outline

  • Boundaries: which tasks are tool-calling vs. pure generation.
  • Structured outputs and JSON schema discipline for downstream systems.
  • Idempotency and partial failure when calling internal APIs.

labs

  • Diagram a target workflow with rollback arrows annotated.

beyond-catalog topics (custom)

  • Sidecar vs. monolithic orchestrator tradeoffs common in regulated enterprises.
  • Runtime policy injection (e.g., per-department tool allowlists).

MCP in the enterprise: rollout & governance

So MCP servers don’t become shadow IT.

session outline

  • Publishing lifecycle for internal MCP servers; versioning expectations.
  • Logging: what must be retained for security reviews vs. what harms privacy if over-logged.

labs

  • Write an MCP server ‘definition of done’ checklist for your environment.

beyond-catalog topics (custom)

  • Secrets hygiene patterns when developer laptops connect to shared sandboxes.

quick contact

Scope or pilot this curriculum

Share sponsor, headcount, and cities — we reply with timing and options. Rough budget helps us match the right depth.

related on-demand courses

faq

Do we need engineers in the room?

Yes for half the modules—otherwise decisions drift into slide-level architecture.

← All curriculum samples·training hub