I build small, sharp tools for the AI-agent era — things I needed myself, shipped when they work, hardened before anyone asks.
| Tool | What it does | Install |
|---|---|---|
| agentport | One config for every coding agent — converts AGENTS.md / CLAUDE.md / Cursor rules / MCP configs bidirectionally. Zero dependencies. | pip install agentport-cli |
| docsqueeze | Feeds PDFs & docs to AI agents inside a token budget — 30-page PDF at ~4k tokens instead of 45k. Anchored, citable output. | pip install docsqueeze |
| alphaverdict | Adversarial fact-checking for LLM-written trading strategies — five reviewers try to falsify your backtest before the market does. | pip install alphaverdict |
- Security-first: agentport survived ~530 adversarial checks before launch
- Zero-dependency where possible — auditability over convenience
- Every claim in every README is measured or testable
Issues are open on all three repos — feature requests and "you did X wrong" both welcome. Building in public: @omrgpt.