Minns Memory Layer

The graph-native database built for agentic AI workflows. Store events, form memories, extract knowledge, and learn strategies — automatically.

Minns Memory Layer is a purpose-built database for agentic workflows. Unlike traditional databases, it understands the intent (Goals) and context of every interaction your AI agents perform.

Every event you log is automatically woven into a knowledge graph that powers long-term memory, episode detection, claim extraction, and strategy learning.

Core concepts

Understand Events, Episodes, Goals, Memory, and Strategies.

Quickstart

Get up and running in under 5 minutes.

TypeScript SDK

Install the SDK and start logging events immediately.

API Reference

Explore every endpoint with full request and response schemas.

How it works

1. Log events

   Send structured events — actions, observations, messages, reasoning — to the POST /api/events endpoint or via the SDK.

2. Automatic graph construction

   Minns Memory Layer builds a knowledge graph in real time: nodes for events, goals, actions, and contexts — connected by typed edges.

3. Episode detection

   Events sharing the same goal are grouped into Episodes. When a goal reaches progress: 1.0, the episode is completed.

4. Memory & strategy extraction

   Completed episodes become searchable Memories. Successful patterns are extracted as reusable Strategies.

5. Query & recall

   Retrieve memories by context, search claims semantically, or ask the Policy Guide “What should I do next?”

Choose your path

I want to build

Jump straight to the quickstart guide.

I want to understand

Read the core concepts first.

I want to explore

Browse the full API surface.