Building AI Agents in TypeScript
summary
Shows how to build production-ready AI agents with Mastra and TypeScript by wiring agents to real tools (example: get budget versus actual tool), designing multi-step workflows with human-in-loop approval (example: cancel subscription workflow), enforcing structured outputs with schemas and running automated evaluations and scorers that caught failures with surprising metrics like 0.14 and 0.45. Covers memory integrations (local MCPs, MeZero), model fallbacks and retries, observable traces, cost/latency metrics, and an OpenAPI/Swagger surface so a custom frontend or local LLM can replace the demo UI.
AI-written summary · about
chapters (21)
- 0:00Building AI agents intro
- 1:54Why build AI agents
- 3:19Control with Mastra
- 3:40Mastra Studio demo
- 5:03Agents calling tools
- 7:59Human-in-the-loop workflows
- 9:25Tool scheduling workflows
- 10:49Shared tools across agents
- 11:53Workspaces and skills
- 12:33Processors and prompts
- 12:55MCP servers for agents
- 13:40Runtime context handling
- 14:01Evaluating agents with scorers
- 15:04Agent experiments
- 16:53Evaluations prevent failures
- 17:34Tracking agent costs
- 18:16Debugging agent traces
- 19:19Custom UI and deployment
- 20:44Mastra client fallbacks
- 21:27Memory and voice in agents
- 22:09Local LLMs with LM Studio
related talks
01h 51m[Workshop] AI Pipelines and Agents in Pure TypeScript with Mastra.ai
49mAgents at the Application Layer
46mFullstack AI: The Building Blocks for Production Ready AI-Powered Software
45mTaming AI Agents With Output.Ai
58mAdvanced Autonomous AI Agents: Memory, Collaboration and Evolution in Action
30mBuilding Reliable AI Agents: An Open Source Approach To Evaluation and Observability
22mHOW TO BUILD YOUR OWN AI AGENT: THE COMPLETE ARCHITECTURE FOR ENGINEERING TEAMS
16mBuilding Applications with AI Agents