Shepherd

Shepherd (askshepherd.ai) markets a shared memory layer for engineering and research organizations. Its stated purpose is to connect research, hardware, firmware, software and operations into a single memory that a team can trace and that AI agents can act on. According to the company website, work artifacts generated in the normal course of a project — design reviews, experiments, messages and commits — are ingested automatically and become shared context available to both people and agents.

The product is offered as a macOS desktop application (macOS 13 or later) and as a web application. The site advertises connections to a wide range of workplace and engineering systems, including Slack, Granola, Notion, Gmail, Google Drive, Google Docs, Google Calendar, GitHub, Bitbucket, Confluence, schematics, Claude Code, Codex, Cursor, Linear and Stripe. Usage patterns shown on the site include asking a question in Slack about why a hardware revision changed and receiving an answer that traces the decision back across a vibration test, a research proposal and a hardware approval; following a specification in Confluence through to a schematic, firmware and code; and linking a Bitbucket change to its system requirement in Linear. The website states the company is backed by Y Combinator.

Note on sources: several documents retrieved under the name "Shepherd" describe a different, unrelated company — a San Francisco commercial construction insurtech (shepherdinsurance.com, YC Winter 2021, founders Justin Levine and Mo Mahallawy) — and a further source describes an unrelated AI consultancy at aishepherd.dev. Those sources were excluded from this profile.