Movez · Launch Video Breakdown: Hook, Pacing & Motion Design

Scene-by-scene timeline & spoken transcript

1. 0:00 – 0:55

   The Hook

   Introduction to Managed Agents & Core Concepts

   “(No spoken dialogue — Driving electronic rhythm and ambient sound design.)”

   On screen

   What we'll cover 01 Core concepts A quick refresher — what Managed Agents are, the server-side loop, and the three resources. 02 Hands-on workshop Build a working incident-investigator agent end to end — six functions in agent.py. 03 Beyond the basics What else is in the box — vaults, subagents, memory, webhooks — and where to go next.

   Camera

   Static wide shot of presenter on stage with a large screen displaying presentation slides. Occasional close-ups on the slides.

   Motion

   Simple slide transitions, text reveals, and presenter gestures. Focus on clear information delivery.

2. 0:55 – 1:50

   Problem Agitation

   Architectural Shift & Agent Building Blocks

   “(No spoken dialogue — Driving electronic rhythm and ambient sound design.)”

   On screen

   Evolution of interfaces used to build agents Messages API Agent SDK Managed Agents Production agent infrastructure YOU MANAGE YOU MANAGE YOU PROVIDE Agent loop + context management Hosting & scaling Tool execution runtime Session state & observability Session persistence & checkpointing Auth (vaults) + credential vault Hosting, auth, observability Custom tool logic Hosting, scaling, observability Tool runtime & sandbox Custom tool logic (MCP) ANTHROPIC HANDLES Purpose-built harness + context mgmt. Tool runtime & sandbox Session persistence & checkpointing Auth (vaults) + credential vault Hosting, scaling, observability Tokens in / out Tokens in / out Tokens in / out Code w/ Claude

   Camera

   Alternating between wide shots of the presenter and close-ups of the presentation slides, highlighting key diagrams and text.

   Motion

   Animated diagrams illustrating the 'before' and 'now' states of agent architecture, with elements highlighting and fading to draw attention to changes.

3. 1:50 – 10:00

   Product Reveal

   Hands-on Workshop: Building an SRE Agent

   “(No spoken dialogue — Driving electronic rhythm and ambient sound design.)”

   On screen

   Part two Hands-on workshop Laptops open — head to the link below. https://cwc26.short.gy/workshops Code w/ Claude INCIDENT-2277 CRITICAL checkout - p99 latency 10x baseline triggered 2020-04-22 14:32:07 UTC policy: bug-oncall - ack - P99 latency 3,638 ms ↑ +572% ERROR RATE 21.9 % ↑ +21.9 pp DB pool 100 % ↑ +90 pp DURATION 28 min ↑ ongoing AFFECTED SERVICE checkout auth inventory SIGNAL 14:32:11 paged 14:32:14 ack 14:32:07 alert fired checkout.p99_latency_ms > 500 for 5m @oncall-platform

   Camera

   Dynamic screen recordings of a VS Code environment and a web application, interspersed with shots of the presenter coding live.

   Motion

   Rapid cuts between code editor, terminal, and web UI. Code highlighting, cursor movements, and text input are animated to simulate live coding. UI elements animate to show data updates and agent interactions.

4. 10:00 – 20:00

   Feature Teaser

   Deep Dive into Agent Functionality & Architecture

   “(No spoken dialogue — Driving electronic rhythm and ambient sound design.)”

   On screen

   The cloud agent calls a function on your laptop Cloud agent loop running on Anthropic agent.custom_tool_use over the SSE stream you hold open user.custom_tool_result POST the result back Your script handle_tool(name, args) No inbound networking. You hold the pipe open; the request rides it back. Swap json.load("data/metrics.json") for a Datadog client — same wire protocol. tool schema (in agents.create) {"type": "custom", "name": "get_metrics", "input_schema": {"type": "object", "properties": {}}} handle_tool() body def handle_tool(name, args): if name == "get_metrics": return json.load(open("data/metrics.json")) if name == "get_recent_deploys": return json.load(open("data/deploys.json")) if name == "get_diff": return open( "f"data/diffs/{args['sha']}.diff").read() → agent.py : handle_tool() : 7 lines

   Camera

   Focus on detailed diagrams and code snippets, with a static camera on the screen. Occasional presenter shots for context.

   Motion

   Flowchart animations, highlighting of code blocks, and animated text overlays to explain complex interactions between components. Elements slide in and out to reveal information.

5. 20:00 – 2209

   Call to Action

   Beyond the Basics: Advanced Agent Capabilities

   “(No spoken dialogue — Driving electronic rhythm and ambient sound design.)”

   On screen

   Beyond the basics Subagents callable_agents — agents that spawn and coordinate other agents. Memory Persistent agent memory mounted into the container. Outcomes A structured artifact of what the agent produced, separate from events. Vaults Per-user credentials registered once, referenced via vault_ids. MCP servers Attach remote tool servers via mcp_toolsetentries. Webhooks Console → Settings → Webhooks. Fires on session_status_idled. Permission policies always_ask on a tool → loop pauses for user.tool_confirmation. Interrupt Postuser.interrupt mid-run to force the agent to idle. Console agent builder Iterate on agent configs interactively before dropping to the API.

   Camera

   Wide shots of the presenter and the screen, showcasing a grid of advanced features. Zoom-ins on individual feature descriptions.

   Motion

   Grid layout with animated icons and text descriptions for each feature. Highlight animations on individual features as they are discussed. Smooth transitions between features.