Can a LangChain Deep Agent direct a live race broadcast? | Nebius Token Factory

One feed cannot show every driver on track. A deep agent opens a player for each, but not on the first try. LangChain's Deep Agents ship a harness on top of LangGraph with tool use, context management, and delegation out of the box. Rob Spectre wires one to a video multiviewer over MCP, picks a model from the Nebius Token Factory catalog, and writes a system prompt that turns it into a broadcast director: read live timing, work out who is pushing, open a player for each of them, tile them, keep them synced. The first run opens players for drivers who are not on a flying lap and syncs them late. Rather than filling the notebook with print statements, Rob adds LangSmith and reads the traces, which show the model spending so long on telemetry that laps finish before the players catch up. The fix is a smaller, faster model from the same catalog. The tiling still needs work at the end. If you want to build like this, come check out the Nebius AI Builder Program. It bundles credits for Nebius Token Factory, Tavily, and Nebius Academy, plus training, office hours, and access to the builder community. Register once at dev.nebius.com/builders. Links: Builder Program: https://dev.nebius.com/builders Token Factory: https://dev.nebius.com/token-factory Chapters 0:00 A bundle of free AI services 0:24 LangChain Deep Agents 0:49 Why one feed misses the action 1:21 Multiviewer, MCP, and dependencies 4:22 Choosing a model on Token Factory 5:47 Writing the system prompt 9:28 Creating the agent and running it 12:19 Debugging traces in LangSmith 16:09 Switching to a faster model