Introduction
beginnerEngramma is a cognitive memory engine for AI agents. Not a vector database — a thinking memory that reasons, consolidates, and explains.
What is Engramma?
Engramma Memory Cloud is a cognitive memory engine designed for AI agents. Unlike vector databases that simply store and retrieve embeddings, Engramma processes memories through a 10-phase cognitive cycle that mirrors how biological memory works.
The result: your AI agents don't just search — they remember, reason, and learn.
Why not a vector database?
Vector databases solve similarity search. Engramma solves memory — a fundamentally different problem:
| Capability | Vector DB | Engramma |
|---|---|---|
| Store & retrieve | Yes | Yes |
| Causal reasoning | No | Yes |
| Memory consolidation | No | Yes |
| Explainability | No | Yes |
| Adaptive retrieval | No | Yes |
| Memory decay & strengthening | No | Yes |
| Regime detection | No | Yes |
Think of it this way: a vector database is a filing cabinet. Engramma is the brain that decides what to keep, what matters, and why something is relevant.
Who is it for?
- AI agent developers who need persistent, reasoning memory
- Chatbot builders who want conversations that truly learn
- Research teams building knowledge systems with causal links
- Enterprise teams who need explainable AI decisions
How it works (30-second version)
- You store text or embeddings via the API
- Engramma routes them through 3 retrieval pathways (Exact, Energy, Attention)
- When you retrieve, the Confidence Router scores across all pathways
- Over time, memories consolidate — strengthening important patterns, pruning noise
- You can ask why any result was returned (explainability)
from engramma_cloud import EngrammaClient
import os
client = EngrammaClient(api_key=os.environ["ENGRAMMA_API_KEY"])
# Store a memory
result = client.store("The user prefers dark mode and concise answers")
print(result.id) # "pat_8f3a..."
print(result.importance) # 0.72
# Retrieve with reasoning
results = client.retrieve("What are the user's preferences?")
print(results[0].text) # "The user prefers dark mode..."
print(results[0].confidence) # 0.94
print(results[0].pathway) # "exact"
# Understand why
explanation = client.explain(query="What are the user's preferences?")
print(explanation.summary) # "High confidence via exact match..."Key concepts
| Concept | What it means |
|---|---|
| Memory Types | Three retrieval pathways — Exact, Energy, Attention — each suited for different queries |
| Causal Reasoning | Ask "what if?" and "why?" — your memories form causal relationships |
| Consolidation | Memories strengthen over time through sleep cycles, just like biological memory |
| Explainability | Every result comes with a reason — understand the confidence scores |
| Regimes | The engine adapts its behavior when it detects novelty or anomalies |
Next steps
- Quickstart — Store and retrieve your first memory in 5 minutes
- How It Works — Understand the 10-phase cognitive cycle
- API Reference — Full endpoint documentation
Already familiar with vector databases? Check the Migration Guide for a direct comparison and step-by-step migration.