Projects
Systems you can inspect.
Working builds, reference architectures, and the tradeoffs hiding behind the tidy diagram.
Build loop
Less magic. More evidence.
- Problem Name the operational risk.
- Boundaries Make policy and constraints explicit.
- Build Create the smallest useful system.
- Evaluate Keep evidence, not vibes.
01
AgentOps Simulator
A simulator for agent workflow cost, routing, guardrails, reliability, and observability.
- Python
- Streamlit
- OpenTelemetry
- Pydantic
- Scenario
- Route
- Execute
- Evaluate
02
AgentDesk
A control-plane exploration for policy-aware computer-use agents.
- TypeScript
- Playwright
- JSON Schema
- Policy engine
- Task
- Policy
- Browser
- Trace
03
EvalKit
A lightweight evaluation framework for LLM and agentic workflows.
- TypeScript
- MDX
- OpenTelemetry
- PostgreSQL
- Cases
- Run
- Score
- Review
04
AI Failure Atlas
A reproducible field guide to how agentic systems fail.
- Python
- Pytest
- OpenAI
- Failure taxonomy
- Failure
- Fixture
- Trace
- Mitigate
05
Secure Tool Gateway
Permissioned tool execution with scoped access, audit logs, and approval gates.
- OAuth
- Policy engine
- TypeScript
- Audit logging
- Request
- Policy
- Approve
- Audit
06
MCP Notes Server
Safe, inspectable MCP access to a local Markdown notes workspace.
- Python
- MCP Python SDK
- Pydantic
- Markdown
- Ask
- Discover
- Validate
- Notes
07
A2A Agent Network
A beginner-friendly network for agent discovery, delegation, and structured task handoffs.
- Python
- FastAPI
- Pydantic
- JSON-RPC
- Request
- Discover
- Delegate
- Return
08
Multi-Agent Planner
A beginner-friendly planner-executor-reviewer workflow for learning multi-agent orchestration without the group chat chaos.
- Python
- Typer
- Pydantic
- Pytest
- Goal
- Plan
- Execute
- Review