chapter nine

9 Building an AI assistant: putting the platform to work

 

This chapter covers

  • Designing a personal AI assistant using our platform services
  • Incrementally adding memory, knowledge, tools, safety, and observability
  • Implementing the agent loop that coordinates reasoning, retrieval, and action
  • Applying context engineering to assemble the right information for every request
  • Deploying the complete application with the Workflow Service

Every platform service we’ve built solves a piece of the AI application puzzle. The Model Service generates responses. The Session Service remembers conversations. The Data Service retrieves organizational knowledge. The Tool Service calls external systems. The Guardrails Service enforces safety policies. The Observability Service tracks what happens. The Experimentation Service measures improvement. The Workflow Service deploys and scales it all. What we haven't done yet is build a complete application that weaves these services together from start to finish. That's what this chapter is about.

9.1 The blueprint

9.2 The simplest assistant: a model call in a workflow

9.3 Teaching Claw to remember conversations

9.4 Long-term memory: remembering across sessions

9.5 Grounding Claw in reality: knowledge retrieval and agentic RAG

9.6 From conversation to action: tools and the agent loop

9.7 Safety: guardrails at every step

9.8 The prompt that makes it all work: context engineering

9.9 The complete agent loop

9.9.1 Streaming the response

9.10 Seeing what Claw does: observability

9.11 Making Claw better: experimentation

9.12 Deploying Claw

9.13 What the platform gave us

9.14 Summary