Introduction

What is Shelvia?

Shelvia captures AI chats, files, sources, decisions, and prompts and turns them into reviewed project memory. Humans can search it. Tools and agents retrieve it through API, MCP, and SDK before they act.

A project is the unit of work. Every memory row lives under a workspace and a project. Reads carry provenance (workspace + project + token + generated_at). Writes always create review candidates, never trusted rows directly.

Who Shelvia is for

• Founders and builders who want their AI work to survive the session.
• AI engineers and agent builders who need a memory layer their tools and agents can call.
• Researchers and analysts who want sources, prompts, and decisions to stay near each other.
• Product and operations teams that need shared context with audit trails.

How to read these docs

• Concepts pages explain the memory model.
• Developer reference pages show how to call the REST API, MCP, SDK, and webhooks.
• Roadmap pages cover planned surfaces (Browser Capture, Agent Compatibility), explicitly labeled.