Video by Rasa via YouTube
Context Engineering for AI Agents: Keep the Record, Send a Focused View with Rod Rivera · Rasa × Prof Rod
Try it yourself: Rasa Mantle Getting Started: https://mantle.rasa.com/getting-started
Chapters:
0:00 Start
0:28 How do AI agents manage context?
2:21 Why not send the model everything?
5:49 What goes into the focused view?
7:15 How does Rasa Mantle manage context?
12:24 Audit any value your agent stores
Audit any value your agent stores:
1. Where is it kept?
2. What selects it?
3. Will the model see it next call?
Rasa: https://rasa.com
Sources used in this video:
Mantle documentation (read September 26, 2026):
Context Management: https://mantle.rasa.com/mantle/context
System Prompt: https://mantle.rasa.com/reference/system-prompt
Execution Loop: https://mantle.rasa.com/reference/execution-loop
Scoped Instructions: https://mantle.rasa.com/build-guide/scoped-instructions
Prompt templates: https://mantle.rasa.com/reference/prompt-templates
Research: Kelly Hong, Anton Troynikov and Jeff Huber, "Context Rot: How Increasing Input Tokens Impacts LLM Performance," Chroma, July 14, 2025: https://www.trychroma.com/research/context-rot
Code on screen is copied from the Mantle pages above; the Inspector screenshot is from a real Mantle run captured September 29, 2026. The token counts on the budget card are illustrative.
Prof Rod: https://profrod.ai
#ContextEngineering #AIAgents #Rasa