Video tutorials
Each video builds something concrete, shows it running, and points to the next step. The code from every tutorial is in sagewai/examples/ so you can run it locally.
Getting started
| # | Title | Duration | What you build |
|---|---|---|---|
| 1 | Sagewai in 5 Minutes: Your First AI Agent | 5 min | Install sagewai, create a 4-line agent with GPT-4o, add a custom tool with @tool, run it. |
| 2 | Run AI Agents for Free with Ollama + Sagewai | 8 min | Install Ollama, pull llama3.1, create an agent with providers.ollama(). No API keys, no cloud costs, completely local. |
| 3 | Sagewai vs LangChain vs CrewAI: Honest Comparison | 12 min | Build the same research agent in all three frameworks. Side-by-side look at cost control, fleet workers, and local inference. |
Core SDK features
| # | Title | Duration | What you build |
|---|---|---|---|
| 4 | Building a Research Agent with Memory | 10 min | An agent that remembers facts across conversations using GraphMemory. The knowledge graph builds as the agent runs. |
| 5 | RAG in 10 Minutes: PDF Q&A with Context Engine | 10 min | Ingest a PDF, ask questions about it, watch retrieval work with the @context directive. Demonstrates vector + graph hybrid search. |
| 6 | Multi-Agent Workflows: Researcher, Analyst, Writer | 12 min | A 3-agent content pipeline with SequentialAgent. Data flows between agents; execution is visible in the admin panel. |
| 7 | Sagewai Directives: The Prompt Preprocessor | 8 min | @context, @memory, @agent:name(), @wf:name(), /tool, #model. Watch how directives transform a prompt before the LLM call. |
| 8 | Safety and Guardrails: PII, Hallucination, Budget Control | 10 min | Add PIIGuard, HallucinationGuard, and TokenBudgetGuard to an agent. See blocked responses, redacted PII, and budget enforcement in action. |
Developer tools
| # | Title | Duration | What you build |
|---|---|---|---|
| 9 | VS Code Extension: Faster Agent Development | 6 min | Install the Sagewai VS Code extension. Directive syntax highlighting, scaffold commands (New Agent, New Workflow, Add Tool), code snippets. |
| 10 | Using Sagewai from TypeScript, Go, and Rust | 10 min | Start the harness proxy, connect from three languages using client wrappers. Same agent infrastructure, three ecosystems. |
| 11 | MCP Server: Give Claude Code and Cursor Your Agent's Tools | 8 min | Create an MCP server with sagewai mcp serve, connect it to Claude Code and Cursor. AI coding assistants call your custom tools. |
Enterprise and production
| # | Title | Duration | What you build |
|---|---|---|---|
| 12 | Fleet Architecture: Server + Workers Explained | 15 min | Deploy a server container, create enrollment keys, start 3 workers (CPU + GPU), submit tasks, and watch pool/label routing. |
| 13 | Multi-Tenant Setup: Per-Team Isolation and Budgets | 10 min | Create 3 projects with different budgets and worker pools. Each team's spend, agents, and data stay isolated on the same server. |
| 14 | Zero-Cost AI: Fine-Tune with Unsloth, Serve with Ollama | 15 min | Generate training data with GPT-4o, fine-tune with Unsloth (4-bit QLoRA), export to GGUF, serve via Ollama. Harness auto-discovers it; $0/token inference. |
| 15 | Production Kubernetes Deployment | 12 min | Deploy Sagewai fleet on K8s: server Deployment, GPU worker DaemonSet with NVIDIA tolerations, HPA for CPU workers. Terraform and Pulumi templates. Monitoring with Grafana. |
| 16 | CI/CD Agents: PR Summaries and Eval Quality Gates | 10 min | Add run-agent and run-evals GitHub Actions to a real repo. A PR summary bot comments on every PR; a quality gate fails CI when agent performance drops. |
| 17 | The LLM Harness: Proxy, Route, and Budget All AI Calls | 10 min | Point Claude Code and Cursor at the harness proxy. Complexity-based routing (Haiku for typos, Opus for architecture), per-team budgets, full audit trail. |
Advanced
| # | Title | Duration | What you build |
|---|---|---|---|
| 18 | Durable Workflows: Temporal-like Execution in Python | 12 min | A workflow with ApprovalGate for human-in-the-loop review. PostgreSQL-backed checkpointing, worker crash recovery, and the dead letter queue. |
| 19 | Self-Learning Agents: Improve Automatically from Usage | 10 min | Enable auto-learn on an agent, watch it store successful interaction patterns. Response quality improves as the agent accumulates runs. |
| 20 | Building a Full-Stack AI App with Sagewai | 20 min | Python backend with FastAPI + Sagewai agents, Next.js frontend, admin panel monitoring everything. End-to-end from code to production. |
Recording structure
Each video follows the same format:
- What we're building (30 seconds) — show the end result first
- Prerequisites (30 seconds) — what you need installed
- Live coding (main content) — build it step by step
- Run and demo (1-2 minutes) — show it working
- What's next (15 seconds) — link to the next video
All code is in the sagewai/examples/ directory of the SDK.