atomicstrata

atomicstrata / atomicmemory

Public

Portable semantic memory for AI agents: core engine, TypeScript SDK, framework adapters, MCP server, CLI, and host plugins.

docs.atomicstrata.ai agent-infrastructure agent-memory ai-agents ai-memory cli
18
1
89% credibility
Found May 27, 2026 at 18 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

AtomicMemory is an open-source memory system for AI applications that gives AI assistants the ability to remember important details across conversations. It works by capturing context from conversations, storing facts persistently, and retrieving relevant memories when needed. The system supports automatic memory updates and corrections, so when users change their mind or provide new information, the AI stays current. AtomicMemory provides ready-made connectors for popular AI development frameworks and can run either locally on your own infrastructure or through a hosted service, giving developers flexibility in how they handle sensitive conversation data.

How It Works

1
💡 You realize your AI needs a memory

Your AI assistant keeps forgetting important details between conversations, and you wish it could remember user preferences and past decisions.

2
📦 You install AtomicMemory

You add the AtomicMemory package to your project using a simple install command, just like adding any other tool to your AI app.

3
🔌 You connect it to your AI framework

AtomicMemory provides ready-made connectors for popular AI tools like Vercel AI, LangChain, and others. You pick the one that matches your setup and plug it in.

4
You choose where your memories live
🖥️
Local setup

Run everything on your own machine - your conversations never leave your computer.

☁️
Hosted service

Let AtomicMemory handle the infrastructure so you can focus on building your app.

5
💬 Your AI starts remembering

As your app runs, AtomicMemory quietly captures important facts from conversations and stores them automatically.

6
🔍 Your AI finds what it needs

When your AI needs context, it searches through past conversations and retrieves the most relevant memories to inform its responses.

✨ Your AI becomes genuinely helpful

Your AI remembers user preferences, past decisions, and important context - making every conversation more personalized and useful.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 18 to 18 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is atomicmemory?

AtomicMemory is a portable semantic memory layer for AI agents built in TypeScript. It gives your AI applications the ability to remember user preferences, past decisions, and conversation context across sessions. The project ships a core engine, TypeScript SDK, CLI, MCP server, and pre-built adapters for popular frameworks like LangChain, LangGraph, Mastra, Vercel AI SDK, and OpenAI Agents SDK. You can run the memory backend locally for privacy-sensitive workloads or use the hosted option when convenience matters more.

The core value proposition is simple: instead of re-implementing memory capture and retrieval for every AI project, you plug in one memory layer and get semantic search, memory mutation (add/update/delete/no-op), and context packaging across all your agent surfaces.

Why is it gaining traction?

The project positions itself against the "black box problem" in AI memory products. Most hosted memory services are opaque systems where you trust them with the layer that decides what an AI believes about your users. AtomicMemory takes the opposite stance: the engine is open source and self-hostable, and the interface is designed to be portable across frameworks.

The correction-aware design stands out. Real applications need more than append-only recall. When users change their mind or facts evolve, AtomicMemory handles supersession and revision rather than accumulating contradictory memories. The benchmark results on BEAM and LoCoMo datasets show competitive performance/cost positioning against published competitors like Mem0.

Who should use this?

AI application developers building agents that need persistent user context. If you're using LangGraph, Mastra, Vercel AI SDK, or OpenAI Agents SDK and want your agents to remember preferences across sessions without rebuilding memory infrastructure from scratch, this is worth evaluating. Privacy-conscious teams who want local-first deployment will find the self-hostable core engine compelling. The CLI and MCP server also make it useful for teams integrating memory into Claude Code, OpenClaw, or Hermes workflows.

Verdict

AtomicMemory solves a real problem with a well-thought-out architecture, but the 18-star count signals early-stage software. The credibility score of 0.8999999761581421% reflects this maturity level. Documentation is solid and test coverage appears thorough, but you're adopting into an evolving ecosystem. The framework adapter coverage is impressive for a project this size. Worth trying for pilot projects if you need cross-framework agent memory, but plan for active monitoring as the API surface matures toward 1.0.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.