KYB: Shared Operational Memory for AI Agents

A 12-slide technical introduction to KYB (Know Your Business), a Git-backed knowledge base, incident tracker and task store for AI agents and people. The presentation explains how operational records persist between sessions, how Markdown files and Git commits provide the canonical record, and how the service rebuilds its search index from that history. Topics include the three record kinds, optional semantic retrieval combined with BM25 through reciprocal rank fusion, the distinction between retraction and archiving completed work, default and historical search, task deadline metadata, and CLI, HTTP API, MCP and web-console interfaces. It also describes deployment and access boundaries: authentication is off by default, tokens distinguish permissions, and external access needs an appropriate trusted network or TLS setup. The route-change and task examples are synthetic illustrations. No private operational data, new benchmark, measured latency improvement or claim of better agent performance is presented. Git history preserves changes but does not independently certify the truth or authorship of direct repository edits. The content is based on KYB v0.3.3, commit 06f63b5ae0f8fdff24abb3b2c1c848ff41758faa. The related software archive has Version DOI 10.5281/zenodo.23683977 and Concept DOI 10.5281/zenodo.22850723. This deposit is a separate presentation record and does not replace the software archive or represent a scholarly paper or a conference appearance. The PDF contains selectable text, vector diagrams and clickable source/project links. The JSON companion records the slide content, layout and source links. Project: https://github.com/alex09x/kyb. Website: https://kybmemory.com. Author: Alexander Panasenko, https://prod.codes/about/.