Problem

Technical teams spread critical context across research notes, design reviews, chat, issue trackers, commits, experiments, and operational systems. As work moves across research, hardware, firmware, software, and operations, the rationale behind decisions becomes difficult to find and verify. The resulting economic pain is wasted engineering time, duplicated work, slower handoffs, and decisions made without the full technical history. Shepherd’s core use case is reconstructing that history—for example, tracing why a sensor mount changed after a vibration test, how research proposed a new geometry, and when hardware approved it—then carrying that context into implementation.

Product / Service

Shepherd is a shared company-memory and technical-context layer delivered through a Mac desktop app and web app. It connects authorized sources such as Google Workspace, Slack, Notion, GitHub, messaging, meeting and office-audio data, coding-agent session metadata, and MCP clients; it then indexes, summarizes, retrieves, and reasons over the resulting material. The product is intended to turn design reviews, experiments, messages, and commits into traceable memory rather than isolated records.

Users can ask questions in Slack and use MCP to provide full team context to Codex, Claude Code, and Cursor. The benefit is cross-domain traceability: people can follow a requirement into a schematic, firmware change, or code commit, while agents can act with the surrounding technical rationale instead of only the local contents of a repository or document. The public site shows both Mac and web access, currently labeled v0.1.158 for macOS 13+.

Market

Shepherd competes in the intersection of enterprise knowledge management, enterprise search, and persistent memory/context for AI agents, with a specialized focus on technical organizations. Its differentiation is the attempt to connect the full path from research to physical systems to software and operations, rather than treating knowledge as a static wiki or limiting agent context to a single coding workspace. Adjacent alternatives include broad enterprise-search and knowledge-management products such as Glean and Guru; coding tools such as Codex, Claude Code, and Cursor are better understood as integration surfaces in Shepherd’s product than as direct competitors.

The available evidence indicates an early-stage company rather than a scaled commercial vendor. Y Combinator lists Shepherd as an active Summer 2026 company founded in 2026, and the company identifies itself as YC-backed while offering a working Mac/web product. Its terms reference possible paid features, pilots, and enterprise deployments, but the available public evidence does not disclose customers, revenue, or paid usage metrics. Shepherd is therefore best characterized as pre-scale, with public traction limited to an active YC-backed product launch and early platform availability rather than demonstrated commercial traction.