Typed model boundary
Build a structured model call with validation and fallback.
Output: Schema + test cases
For builders, AI leads, and technical teams
Agent Building
Build agents that survive permissions, failures, review, and real operating cost.
Progress from workflows and structured model calls to tools, memory, RAG, MCP, orchestration, human control, evaluation, and production deployment.
Open module 00- Outcome
- A reviewed marketing research agent with traces, evals, and a production readiness decision
- Estimated effort
- 12โ16 hours plus a 3โ5 hour capstone
- Prerequisite
- Comfort with HTTP, JSON, one programming language, and basic application architecture.
Beginner โ production
The ordered field path.
Each module answers one search intent, produces an operating artifact, and links to the previous and next decision. Follow the sequence or enter where your current workflow is stuck.
- 00AI Agents vs WorkflowsChoose a deterministic workflow or an agent based on uncertainty, not fashion.35 min
- 01LLM APIs & Structured OutputsTreat model calls as typed, observable system boundaries instead of free-form chat messages.50 min
- 02Tool Use & Function CallingExpose narrow, typed capabilities so a model can request actions without owning execution policy.55 min
- 03Context, Memory & StateSeparate what the model sees now, what the application knows, and what should persist across runs.55 min
- 04RAG & Knowledge RetrievalRetrieve authoritative, permission-aware evidence before asking a model to answer or act.65 min
- 05MCP & External SystemsUse MCP to standardize model access to tools and context without confusing interoperability with security.60 min
- 06Agent Loops & OrchestrationControl the observe-decide-act loop with explicit state, budgets, and termination conditions.60 min
- 07Multi-Agent PatternsSplit work across agents only when boundaries improve context, ownership, evaluation, or parallelism.60 min
- 08Human-in-the-Loop & GuardrailsPlace people where judgment changes risk, and give them evidence they can review quickly.55 min
- 09Evals, Observability & Failure ModesEvaluate task behavior before launch and observe complete runs after deployment.65 min
- 10Production Deployment, Security & CostOperate agents as production software with least privilege, failure recovery, version control, and full run economics.70 min
- 11Capstone: Build a Marketing Research AgentCombine retrieval, tools, state, human review, evaluation, and production controls in one useful business workflow.3โ5 hours
Hands-on projects
Build evidence of operating ability.
Permission-aware retrieval
Ground answers in evidence without crossing user access boundaries.
Output: RAG flow + access tests
Marketing research agent
Combine tools, state, review, evals, and cost controls in one capstone.
Output: Working agent + readiness review
Tools, field guides, and services
Continue through the graph.
Tool
AI Model Leaderboard
Compare current model capabilities and choose a test set for your use case.
Open resourceField guideDocument-level permissions in RAG
A production guide to permission-aware retrieval and caching.
Open resourceServiceAgentic Workflow Delivery
See how Tenten turns a bounded workflow into an operated AI system.
Open resourceFrom capstone to production
A working demo is the beginning of the engineering problem.
Tenten helps teams scope one agentic workflow, integrate real data and permissions, define evals and handoffs, and own the path through launch and on-call.