Shepherd
Shepherd is a company-memory service that ingests messages, documents, meetings, and code-session metadata from tools like Slack, GitHub, and Notion into a searchable, structured knowledge base. Teams can ask natural-language questions to trace decisions across hardware, firmware, and software domains, and AI agents can draw on the same context. The platform is accessible via dashboard, desktop app, CLI, and MCP tools.
- Enterprise Software
- Artificial Intelligence
- Developer Tools
- AI Agents
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Engineering and research teams working across hardware, firmware, software, and operations often lose critical context as decisions, experiments, and communications are scattered across separate tools like Slack, GitHub, Notion, and email. This fragmentation makes it difficult to trace why a design decision was made, how a change in one discipline affects another, or what context an AI agent needs to act correctly.
Solution
Shepherd creates a shared company memory that continuously ingests messages, documents, meetings, code-session metadata, and source metadata from connected tools, then structures it into searchable wiki artifacts and embeddings. The platform lets users ask natural-language questions—such as why a hardware revision changed a sensor mount—and receive answers grounded in the actual decision trail across research, hardware, firmware, and software. It also provides agents with the same contextual memory, enabling them to act on up-to-date project knowledge. The service is accessible through a dashboard, desktop app, CLI, and MCP tools, with retrieval, summarization, and reasoning handled by automated systems.
Target Audience
Primary users are engineering and research teams at hardware and software companies that need to preserve cross-disciplinary decision context, as well as organizations deploying AI agents that require shared project memory.
Features
- Connects to Google Workspace, Slack, Notion, Granola, GitHub, Discord, WhatsApp, Stripe, and MCP clients to ingest authorized content and metadata
- Local macOS collector reads Codex and Claude Code session records, redacts sensitive values locally, and uploads only structured messages and metadata
- Natural-language query interface that traces decisions across research, hardware, firmware, and software domains
- Wiki artifact generation and embeddings that turn raw communications and documents into searchable, structured memory
- CLI and MCP tooling for developer workflows, plus SSO authentication via WorkOS
- Scoped access controls, audit paths, and support-access boundaries designed to protect customer data