Experiment
Agent Playground
A minimal sandbox for testing AI agent workflows before wiring them into production.
Agent Playground is a React Native shell for prototyping tool-calling flows without committing to a full backend.
Features
- Step-through view of each agent turn
- Mock tools for calendar, search, and file reads
- Export conversation traces as JSON for debugging
Why it exists
Clients kept asking about agents before they had a clear use case. This project gives us a shared surface to explore ideas on a call instead of debating abstractions.
Current state
Still an experiment — intentionally small. The value is in the workflow, not the polish.