mhn-ai-agent-memory

<div align="center">

[](https://www.python.org/downloads/) [](LICENSE) [](#development) [](https://arxiv.org/abs/2008.02217)

Give your AI agents real memory. Not a database with an LLM wrapper. Actual associative memory backed by mathematics.

Install &#8226; Quick Start &#8226; Features &#8226; How It Works &#8226; API Reference

</div>

---

<div align="center">

          ┌─────────────────────────────────────────────┐
          │                                             │
          │   xi_new = X @ softmax( beta * X^T @ xi )  │
          │                                             │
          │   One equation. One matrix multiply.        │
          │   Deterministic. Microseconds. Free.        │
          │                                             │
          └─────────────────────────────────────────────┘

</div>

---

When an AI agent needs to "remember" something today, the standard approach is:

1. Store text in a database
2. Call an LLM to reason about it (costs money, takes seconds, non-deterministic)
3. Hope the LLM doesn't hallucinate about what it stored

This library replaces that pipeline with the Modern Hopfield Network update rule -- the same mathematical structure as transformer attention, but exposed as an explicit, controllable memory.

| | Traditional (LLM + DB) | This Library | |---|---|---| | Retrieval | LLM API call | One matrix multiply | | Latency | Seconds | Microseconds | | Cost |