Building Apps & Multi-Agent Systems
Ship real AI agents and multi-agent systems that work in production.
A hands-on engineering track where you build, evaluate and deploy genuine LLM applications and multi-agent systems — from RAG and tool-calling through graph-based orchestration, the Model Context Protocol, and full production deployment with observability, using the same frameworks teams ship in 2025–2026.
- 60 hours
- 22 modules
- Cohort-based · flexible batches
Pricing is set per cohort and organisation — Talk to us.
Outcomes
What you'll be able to do
Concrete, demonstrable skills — the kind you can show in an interview or put to work on day one.
Architect and build production-grade LLM applications with RAG, tools and memory
Implement multi-agent systems in the leading 2025–2026 frameworks and choose between them defensibly
Author and consume MCP servers to make agents composable and reusable
Evaluate agent quality with offline eval suites, LLM-as-judge and tracing
Deploy, monitor and guardrail an agent as a real streaming API
Curriculum
The modules
Sequenced to build on each other. Each module is a first-class unit — the same ones our AI recommender draws on to map a personalised track.
22
modules
83
hours total
LLM Application Foundations
Build your first LLM app end-to-end — prompt management, structured output, streaming and the provider landscape (OpenAI, Anthropic, Google, open-weight) — and when to pick the Vercel AI SDK, LangChain or LlamaIndex.
Understanding Local LLMs and Why They Matter
What LLMs are and how they work in plain language, and the practical difference between open-source and closed commercial models.
Power Features — RAG, Local Server, and MCP
LM Studio's advanced capabilities — RAG for document-grounded AI, local API server setup, and MCP integration for connecting external tool ecosystems.
Retrieval-Augmented Generation & Vector Search
Engineer real RAG pipelines — chunking, embeddings, hybrid search, re-ranking and citation — and choose between managed vector stores and Postgres + pgvector, measuring quality with RAGAS.
Tool Calling & Function-Calling Agents
Give models real capabilities — define tools, run function-calling loops, manage state and safely execute code and API calls — building one agent that browses the web, runs Python and writes to a database.
Multi-Agent Orchestration Frameworks
Build hands-on in LangGraph, CrewAI, AutoGen/AG2, the OpenAI Agents SDK and Google ADK — supervisor, peer-to-peer and hierarchical topologies — then choose the right framework by control vs autonomy vs ergonomics.
MCP Servers: Build & Consume
Author a custom MCP server exposing your own API, and connect external servers (GitHub, Postgres, file system, Stripe) to a multi-agent system — the standardized, reusable tool layer for any agent.
Getting Started with Antigravity
Introduces the Antigravity platform, its purpose, and the agent-first development philosophy. Covers installation, authentication, and initial environment configuration.
Core Interface and Agent Tools
Explores the Agent Manager, Editor Workspace, browser integration, source control, prebuilt skills and the feedback system. Learners navigate the platform and complete hands-on tasks using its core components.
Advanced Features and Security
Covers command security controls, the Model Context Protocol (MCP), and integration with external systems such as GitHub MCP and Stitch MCP. Learners build secure, extended agent workflows.
Introduction to AI App Builders
The landscape of no-code AI development tools — Lovable, V0 and Z.ai. How prompt-driven development works and how to generate and refine app ideas with AI.
Building, Debugging, and Securing Applications
Building complete apps through prompt-driven development, debugging efficiently, and implementing security best practices before deployment.
Git and Repository Foundations
The difference between Git and GitHub, installing and configuring Git, the repository lifecycle, essential commands, and creating a first repo with a professional README.
Branching, Merging, and Pull Requests
Structured branching, merge strategies, conflict resolution and the full pull request process. Learners implement GitHub Flow and use Issues to collaborate.
OpenClaw Architecture and Core Concepts
What AI models, agents and proactive systems are in the OpenClaw context — installation methods, remote control from mobile, and proactive application workflows.
Building End-to-End Automation Systems
The full application development workflow — connecting and managing complete workspaces, designing proactive systems and deploying end-to-end automation.
Environment Setup and Model Interaction
System requirements, installing Python, GPU drivers and Unsloth, and Hugging Face integration. Learners load a model, understand tokenization and run first interactions with chat templates.
Data Preparation and Fine-Tuning
Training dataset structure and quality, and parameter-efficient fine-tuning with LoRA — training configuration, resource optimization, progress monitoring and checkpoints.
Human-in-the-Loop, Memory & State
Add persistence, long-term memory and approval gates — interrupt/resume patterns, checkpointing and human approval of destructive tool calls before execution.
Evaluation & Observability
Stop guessing if your agent works — instrument traces, build offline eval suites, run LLM-as-judge and track token cost and latency across the leading observability stacks.
Deployment, Guardrails & Production Engineering
Ship the agent as a real service — containerize, expose streaming APIs, add auth, rate limits and input/output guardrails, prompt CI with eval gates, and deploy in Python (FastAPI) and TypeScript (Vercel).
Capstone: Ship a Real Multi-Agent Product
Design, build, evaluate and deploy a complete multi-agent system of your choice — a research analyst, coding teammate or customer-ops agent — presented with live traces and an eval report.
Questions
Good to know
Still wondering about something? Ask us directly in the enquiry form below.
Do I need to be a strong coder?
Comfortable Python is enough; TypeScript is a plus. Starter repos and templates provided so you focus on agent logic, not boilerplate.
Which framework will we use most?
You'll build in at least three (LangGraph, CrewAI, OpenAI Agents SDK) to understand trade-offs, then go deep on the one that fits your capstone.
Enquire
Talk to us — we'll help you choose.
Tell us where you're starting from and what you want to build. We'll walk you through the cohort, the lab, and whether this is the right first step — or point you somewhere better.
- A real reply from our team — never a bot wall.
- Honest guidance on fit, prerequisites and timing.
- Cohort dates and how the hands-on lab time works.
Keep exploring
Programs you might pair this with
Ready to startbuilding?
Enquire about this course, book a demo, or let our AI map your perfect track. The lab is open.