Skip to content

overview.mdx

Latest commit

 

History

History
73 lines (53 loc) · 4.23 KB

overview.mdx

File metadata and controls

73 lines (53 loc) · 4.23 KB
Raw
title Overview
description MemoryLake is an AI memory platform — store, compute, and retrieve long-term memory for people, teams, and AI agents

What is MemoryLake?

MemoryLake gives AI a memory that lasts. Upload documents, hold conversations, and connect your AI tools — MemoryLake extracts durable memories from all of it, keeps them consistent, and serves them back to any model or agent that needs context.

Every LLM call is stateless by default. MemoryLake adds the missing layer: a shared, queryable memory pool that persists across sessions, apps, models, and teammates.

The Platform at a Glance

The core product: workspaces, projects, documents, and extracted memories — with a Playground to experience memory-augmented chat and integrations to bring memory into your tools. A drop-in gateway that adds persistent memory to your existing OpenAI / Anthropic API calls with a base-URL change. Private preview. A unified, OpenAI-compatible entry point to mainstream models — one API key, one endpoint, native paths preserved for Claude and Gemini. Shared workspaces with roles, fine-grained permissions, member quotas, and usage analytics — memory that belongs to the organization, not just one person.

Core Capabilities

  • Memory extraction: Conversations and documents are distilled into structured, deduplicated memories — not just stored as raw text
  • Memory retrieval: Hybrid semantic + keyword search across memories and document content, scoped by workspace and project
  • Memory hygiene: Source traces show where each memory came from; conflict detection surfaces contradictory memories for you to resolve
  • Multimodal documents: PDF, Word, Excel, PowerPoint, Markdown, text, and images are parsed and indexed for retrieval
  • Open Data: Subscribe projects to curated industry datasets (research, clinical, financial, patents, and more) that agents can search alongside your own content
  • Integrations: OpenClaw / QClaw / Hermes agent plugins, a REST API, and an MCP server — memory follows your agents wherever they run

Individual Memory and Organizational Memory

MemoryLake is built around a simple idea: memory has an owner and a scope.

  • Personal: your own workspace remembers your preferences, decisions, and files — every AI tool you connect gets the same context about you.
  • Organizational: team workspaces accumulate shared knowledge — project documents, decisions, and domain facts — governed by roles and per-resource permissions, so institutional memory survives personnel changes.

Learn how the pieces fit together in Core Concepts, or jump straight into a scenario:

Give every AI tool you use one consistent, private memory of you. Turn scattered documents and decisions into governed organizational memory. Make OpenClaw and other coding agents remember context across sessions. Add per-user long-term memory to your own product via API or Memory Router.

Getting Started

  1. Sign up — a default workspace is provisioned for you automatically.
  2. Try the Playground — tell it something to remember, add Open Data, upload a file, and watch memory work.
  3. Create an API key — one key authenticates the REST API, Model Router, and agent plugins.
  4. Connect your tools — install the OpenClaw plugin or integrate via API so memory follows you everywhere.