Hi, I’m Melanie nyghtowl owl logo I build and demo AI agents with durable execution at Temporal. Find me on LinkedIn and GitHub. Temporal wrote about my path from film school to AI.

Start here · Temporal & AI · Demos · Vibe Check · Resources

Latest Temporal & AI: AI Agent Gateway & Nexus

Latest Vibe Check: Agent Harnesses: What, Why, and When Not

Start here

Temporal is an open-source platform for durable execution. Your code survives crashes, outages and restarts, which makes it a good fit for AI agents running long, multi-step, failure-prone work like LLM calls, tool use and human-in-the-loop approvals.

  • Install the CLI: brew install temporal then temporal server start-dev
  • Add the agent skills to any AI coding agent (Claude Code, Codex, Cline): npx skills add https://github.com/temporalio/skill-temporal-developer
  • Docs and downloads: temporal.io

Temporal & AI

Temporal & AI is my new program at Temporal, a video series on building AI agents. Every episode has a video and a demo you can run. Here’s the latest, plus the first one, which I made.

AI Agent Gateway & Nexus

Dustin Sweet, a Senior Solutions Architect at Temporal, puts Claude Code behind an agent gateway so every tool call goes through one place you can see and control. A production deploy waits in a Workflow until the right person signs off, and each team shares its own tools through Temporal Nexus without touching the gateway’s code.

Temporal & AI: AI Agent Gateway & Nexus

Agent Memory & State (episode 1)

The same autonomous refund loop run twice. The naive Worker dies before calling Stripe and restarts from scratch, while the durable one reconnects to its Temporal Workflow and picks up at issue refund. No agent framework involved, which is the point.

Temporal & AI: Agent Memory & State

Demos

Durable Research Fleet · Rust Confessional · Durable HITL Agents · ADK & Temporal · Sheep Audio Dreams

Durable Research Fleet A deep-research agent on Temporal Serverless Workers that fans a question out to three to six parallel researchers on Cloud Run and merges one cited report. The worker pool scales to zero while the Workflow stays parked waiting on your approval, then wakes when you give it.

Originally Shubham Londhe’s project. I moved it to Gemini and we demoed it together at a GDE talk.

Rust Confessional

A durable AI agent in Rust, live on stage: the audience confesses programming sins and Ferris the agent delivers judgment, keeping its place through rate limits, network partitions and a Worker crashed mid-absolution. Wall of Regrets is the same demo running as a passive booth display.

Co-presented with Melissa Herrera.

A 10-second walkthrough: the judgment parks at its human-in-the-loop checkpoint, the Worker is replaced mid-deployment, the reply arrives during the gap, and a fresh Worker finishes the same agent.

Durable Human-in-the-Loop Agents

Framework-agnostic human-in-the-loop patterns: agents that pause for human input and reliably pick back up, no matter how long the wait or which framework you use.

ADK & Temporal Integration

Google Agent Development Kit integrated with Temporal for durable, production-ready AI agents.

Voice Demo: Sheep Audio Dreams

A voice-driven AI agent demo built with Temporal.

Vibe Check Series

My ongoing YouTube series exploring AI tools, agentic coding, and how to build AI in production. Watch the full Vibe Check playlist.

Latest episode: Tom Wheeler co-hosted with me for the first time, with guests Jason Steving and Alex Tideman from Temporal. We talk through the open source Temporal Agent Harness (blog post).

Vibe Check: Agent Harnesses: What, Why, and When Not

Resources