Comparison

Runlet is not trying to replace every agent framework. It is a small runtime core for applications that need explicit control over execution, hooks, context budgeting, and observability.

This page explains positioning. For actual usage, start with README.md and the getting-started guides.

OpenAI Agents SDK

The OpenAI Agents SDK has strong lifecycle and tracing ideas. Runlet should borrow the clarity of runner-style execution and first-class tracing, but remain provider-neutral at the core.

Pydantic AI

Pydantic AI has useful type-safety and dependency-injection patterns. Runlet should borrow strong schemas and explicit dependencies without making structured output the central product identity.

smolagents

smolagents shows that agent APIs can stay small. Runlet should keep the default path simple while adding stronger runtime boundaries for hooks, context safety, and observability.

LangGraph

LangGraph is powerful for graph and workflow execution. Runlet should borrow the importance of state and event streams, but should not start as a graph-first workflow engine.