Division two
Agents and Crypto Lab
Where I build and test tools for AI agents and on-chain systems, in the open. The Lab sharpens the same safety skills I bring to small businesses.
Agent spend guard
An open-source guard that lets an AI agent do useful work but stops it before it pays, sends or signs past limits you set. Every decision leaves a receipt a person can check. In development, test networks only.
Agent tooling
Model Context Protocol (MCP) tools that connect agents to real systems with tight permissions and tests.
Agent payments research
Notes on how agents pay for things (such as the x402 standard) and where the security risks are. Research only, on test networks.
Small and honest
The Lab is a small, part-time effort. It ships little, carefully, and says plainly what is unfinished.
Lab rules
- Research and engineering only. Nothing here is investment, financial or tax advice.
- No performance or returns are claimed, and none will be published.
- No tokens. No client money or keys are ever held. Test networks only.
- Lab work is never pitched to small-business clients.
Building an agent?
The Lab is research. It is not a priced service. The service to book is the repair-shop website assistant.