Local execution trees for TypeScript AI agents. agent-inspect helps you understand what happened inside an AI agent run — locally. It turns manual steps, tool calls, LLM calls, structured logs, failures, durations, and run metadata into readable execution trees you can inspect from the terminal. It is built for TypeScript/Node.js developers..

1st Place Winner (General Judge) - Datadog Self-Improving Agents Hack. Two identical AI agents play Split or Steal. No pre-programmed betrayal. They discover deception on their own. Built with @evancorrea.

Braintrust tracing plugin for pi

My implementation for a kaggle competition: https://www.kaggle.com/competitions/WattBot2025

his repository is a sandbox for exploring and building agentic AI applications. It provides a foundation for experimenting with AI-driven solutions, particularly in the domain of decision-making and automation.

A Python implementation of the canonical agent architecture: a while loop with tools. Build production ready AI agents with purpose-built tools, comprehensive tracing, and async patterns.

A causal debugger for distributed AI systems—isolated Daytona worlds, Braintrust evals, and evidence-first incident response.

AI-powered coloring page generator for kids, parents, and teachers.

An eval and observability cockpit for coding agents. It runs policy-controlled coding agents in sandboxed toy repos, tool-use traces, MCP tools, compares harness policies, scores recovery and safety behavior with Python evals, and supports CI-gated verification and Braintrust/Weave export.

LLM evaluation framework for a Singapore personal finance AI agent — Ollama/llama3.2 locally, scored via Braintrust and Langfuse

Can you eval an art form? Canon is a continuity linter for serialized TV, YouTube and micro-drama fiction. Canon plays the role of whats currently the scriptwriting coordinator, verifies your story's logic and cites the scene. No GenAI writing here. That's left to the humans.

A self-improving AI exploration agent built with Google Vertex AI, BigQuery, and Firebase to autonomously analyze data and refine its own querying strategies.

Orchestrate other AI CLIs (agy, Codex, Grok, OpenCode, Claude Code) for second opinions, research, and codebase analysis. No Gemini CLI. Hybrid always-on skill (eval-backed).

coding agent example solving for one-shot runnable applications

Darwin evolves the best whole agent for your task, its prompt, tools, code, and the model it runs on, generation over generation, inside sandboxes it can't escape and against a grader it can't game.

Dual-modular platform combating SRE alert burnout and securing generative AI deployments using Spring Boot and FastAPI. Tracks on-call fairness via Gini coefficients and uses AST/CST analysis to detect fragile, low-quality, or overly AI-dependent code in CI/CD pipelines while introducing quantifiable engineering contribution metric

CacheCatch audits AI agent context and shows what to move so repeated tokens hit cache instead of full-price input — it finds context waste, prompt cache misses, and hidden agent cost leaks; and give you actionable insights how to fix them.

A regulated-agent deployment kit for turning traces, evals, regressions, and approval gates into launch/no-launch decisions

Agent memory patterns: From circular buffers to hierarchical memory systems with summarization and semantic search

Pulse - Track Braintrust/AI engagements and collect testimonials. MCP-enabled for agent integration.