Deesha Tech Academy
Menu

Deesha AI · SkillLab

WorkshopRuns on requestAI & Agentic AIMCPTool callingPermissionsPythonJava

Give Your Agent Hands: MCP Tooling Lab

Two hands-on days connecting an AI application to real systems: tool calling done properly, MCP servers wired in and scoped, safe reads from a live database, business actions gated behind human approval, and the tracing that tells you what the model actually did with the access you gave it.

Designed for

  • Developers whose AI application can talk but cannot act
  • Teams wiring assistants into internal APIs and databases
  • Engineers who configured an MCP server once and want to understand what happened
  • Anyone deciding what an AI application may touch, and how to prove it touched nothing else
Fee
₹2,700
Duration
2 days · 16 hours
Each day
09:00 – 17:00
Mode
Offline / Online
Request a schedule

The 2 days, hour by hour

10 hands-on sessions: every one ends in a thing

Day 1

09:00 – 17:00

Give it hands

5 sessions

09:00 – 10:15

See what tool calling actually is

the tool loop on the wire, schemas as the contract, where MCP fits

Takeaway One tool call read message by message (request, decision, invocation, result) so nothing later is folklore.

10:30 – 11:45

Design tools a model cannot misuse

schema design, naming and descriptions that steer, failing usefully

Takeaway A vague tool redesigned until the model stops misusing it, the improvement measured in call error rate.

12:00 – 13:00

Wire in your first MCP server

connecting existing servers, discovery, scoping what is exposed

Takeaway An off-the-shelf MCP server connected to your application, with half its tools deliberately switched off.

14:00 – 15:15

Let it read a real API

external APIs as tools, auth at the boundary, response shaping

Takeaway The model answering questions it can only answer by calling a live API, keys never entering the prompt.

15:30 – 17:00

Let it read your database, safely

read-only access patterns, query scoping, what must never be exposed

Takeaway Natural-language questions answered from a real database through a scoped read-only tool, and one forbidden table proved unreachable.

Day 2

09:00 – 17:00

Let it act, within limits

5 sessions

09:00 – 10:15

Give it a business action

actions vs reads, side effects behind tools, transaction boundaries

Takeaway The model completing a real business operation through a tool, with the write path wrapped in a transaction it cannot escape.

10:30 – 11:45

Put a human before the consequence

approval gates, what needs approval and what does not, escalation

Takeaway A consequential action paused for explicit approval, and proceeding cleanly after it, demonstrated both ways.

12:00 – 13:00

Hold the security boundary

least privilege per tool, credentials and secrets, injection through tool results

Takeaway A prompt-injection attempt arriving through a tool result, caught at the boundary instead of obeyed.

14:00 – 15:15

Trace everything it does

tool-call tracing, audit logs, cost of tool-heavy conversations

Takeaway A trace that accounts for every tool call in a session: who asked, what ran, what it returned, what it cost.

15:30 – 17:00

Wire the full application and defend it

API plus database plus action together, the access review, demo

Takeaway Your application using all three systems in one conversation: demoed, traced, and its access defended to the room.

By the end of day 2, you are holding

Your AI application connected to three real systems (an API, a database and a business action) through tools and MCP, with permissions scoped to least privilege, a human approval gate on the consequential action, and a trace log that accounts for every call the model made.

No dates announced

We open batches around demand. Tell us the schedule that would work and we will let you know first.

Send your details instead

How will you join?

Laptop ready with the prerequisites?

Opens WhatsApp with a message naming this workshop. We confirm your seat and payment by reply. This website stores nothing.

More SkillLabs

One weekend, one capability

Browse the SkillLabs catalogue

Agent Builder Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

AI App Engineering Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Ground the Model: RAG Engineering Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Idea to Internet

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Agents in the Wild: Production Agent Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Break It Before Users Do: AI Reliability Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Docker & Kubernetes for Developers

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Git, GitHub and GitHub Actions for Developers

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

JavaScript and TypeScript Foundations

5 days · 10 hours

Runs on request

See the plan

Ship It on AWS: Cloud Deployment for Developers

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

SQL for Application Developers

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Events in Motion: Kafka & Messaging for Java Developers

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Linux Command Line Essentials

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Logging & Monitoring for Production

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Performance & Production Readiness

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Redis & Caching for Java Services

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Secure the Frontend: Authentication & Web Security Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Test It Before You Ship It: Java API Testing Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Test the User Journey: Frontend Testing Lab

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

TypeScript for Application Developers

₹2,700 · 2 days · 16 hours

Runs on request

See the plan

Who runs it

Practitioners, in the room with you

All trainers and mentors

Each workshop names its trainer before you book.

  • Vishal Shah

    Vishal Shah

    Founder & Principal Trainer

    Two decades building and teaching commerce, banking and cloud platforms, and still writing code

    • Java & Spring Boot microservices
    • TypeScript, React, Angular & Next.js
    • Composable commerce (commercetools)
    Full profile →
  • Shrenik Shah

    Shrenik Shah

    Principal Trainer

    Cloud and AI architect who has upskilled over 5,000 engineers in Java, React, Python and cloud

    • Python: Django, Flask, GenAI & agentic workflows
    • Java & Spring Boot
    • React & Angular, micro-frontend architecture
    Full profile →

Running this for a team?

We deliver SkillLabs on site for institutions and engineering teams.

Talk to Deesha

₹2,700

2 days · 16 hours

Request a schedule