skip to content

say hello

Building something around AI memory, agents or on-device models? I read everything that lands in my inbox.

End-to-end quality, privacy and compatibility checks for a memory API.

private

An agent-driven acceptance harness that runs realistic memory journeys (sign-up, save, ask) and checks answer quality, tenant isolation, export and delete, MCP tools, the SDK and the browser surfaces.

highlights

  1. 01Scenario and pipeline runner with assertions, checkpoints, a permissions model and automatic secret redaction.
  2. 02Offline fake-API mode for CI (JUnit output) and replayable reports built from event logs. CI is green.

built with

  • node.js
  • playwright
  • mcp
  • github actions
↑↓ move↵ open⌘K toggle