snath-ai/lar
Lár the Pytorch for Agents is the open-source "glass box" engine for building, debugging, and running auditable, self-correcting AI agents.
Building reliable AI agents can be challenging when you can't see what's happening inside. This project helps developers create AI agents that are transparent, predictable, and easy to fix. It takes your agent's code as input and provides a detailed log of every step, allowing you to trace decisions, debug errors, and understand costs. This is for software developers and AI engineers who need to deploy trustworthy, mission-critical AI agents.
Use this if you are a developer building an AI agent for a critical application and need to ensure it's auditable, debuggable, and operates predictably.
Not ideal if you're looking for a low-code platform to quickly prototype simple agents without needing deep visibility or control over their internal execution.
Stars
11
Forks
2
Language
Python
License
Apache-2.0
Category
Last pushed
Feb 27, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/agents/snath-ai/lar"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
openai/openai-agents-python
A lightweight, powerful framework for multi-agent workflows
openagents-org/openagents
OpenAgents - AI Agent Networks for Open Collaboration
vamplabAI/sgr-agent-core
Schema-Guided Reasoning (SGR) has agentic system design created by neuraldeep community
BrainBlend-AI/atomic-agents
Building AI agents, atomically
camel-ai/camel
🐫 CAMEL: The first and the best multi-agent framework. Finding the Scaling Law of Agents....