Shepherd
Shepherd indexes a technical company's everyday tools into one shared memory its people and agents can query.
Shepherd indexes a technical company's everyday tools into one shared memory its people and agents can query.
Shepherd's differentiator is breadth across technical domains: its first-party pages position the memory as spanning research, hardware, firmware, software, and operations rather than a single department's documents, with schematic files treated as a first-class source alongside messaging and code.
A local collector reads Codex and Claude Code session records, redacts sensitive values on the device before upload, and Model Context Protocol tools hand the same indexed context to coding assistants, which is an agent-native path most workspace-search incumbents do not expose first-party. Terms also disclose stricter no-employee-plaintext deployment controls for customers that configure them.
Source: askshepherd.ai
Shepherd is an early-release product from a two-person founding team, which limits delivery depth against enterprise-search incumbents such as Glean, Dust, and Notion AI that publish pricing, documentation, security pages, and changelogs. Every one of those first-party surfaces returns 404 on askshepherd.ai as of September 2026, and the only contact route is a founders' mailbox and a booking link.
Distribution is a direct macOS download plus a web application, with no app-store listing, and the company publishes no newsroom or launch material that a buyer could use to evaluate momentum.
Source: askshepherd.ai
Shepherd is an early-stage YC S26 company operating from San Francisco with a two-person founding team listed on LinkedIn, selling an AI company-memory platform in early release on macOS with a web application and MCP tools. First-party contact is founders@askshepherd.ai and a Cal.com booking page addressed to the founders, and the site's terms and privacy pages were both updated June 28, 2026.
Y Combinator's directory page for the company returned 404 on repeated fetches this run, so the accelerator claim rests on the first-party badge and the LinkedIn company name; a third-party funding database attributes YC partner Diana Hu to the batch. No funding round beyond YC's investment is recorded on any surface walked.
Source:
Shepherd publishes no prices and offers no self-serve plan, seat, or credit pricing.
Sales run through negotiated pilots rather than self-serve billing, while the product itself is free to download as a macOS application with a web application beside it.
Source: askshepherd.ai
Shepherd sells one eponymous offering: an AI company-memory platform that turns the systems a technical organization already runs into a single indexed record. Connected sources named on the first-party pages include Slack, Google Workspace, Notion, Granola, GitHub, Bitbucket, Confluence, Linear, Stripe, office audio, and local Codex and Claude Code session metadata.
Delivery is a macOS desktop application with local source collectors, a web application, and APIs with CLI and Model Context Protocol tools, so coding assistants receive the same team context. The service is in early release, and no public pricing page or documentation site exists as of September 2026.
Source: askshepherd.ai
