Skip to content

CogDB Documentation

A second brain for AI agents — episodic, semantic, and procedural memory in one engine.

Quick install

pip install git+https://github.com/MuLIAICHI/cogdb.git

from cogdb import CognitiveDB

db = CognitiveDB(db_path="./memory")

# Store what happened (episodic)
db.remember("Deployed v2.3 — CORS error on /users", agent_id="devops", importance=0.9)

# Store what's known (semantic)
db.learn("api_service", "version", "v2.3", agent_id="devops", confidence=1.0)

# Store how to fix things (procedural)
db.learn_procedure("fix_cors", steps=[...], agent_id="devops")

# Recall with a token budget
memories = db.recall("CORS error", agent_id="devops", token_budget=500)

What's in here

  • Getting Started

    Build a DevOps assistant agent with full tri-memory support in under 30 lines of Python.

  • API Reference

    Complete reference for every public method — parameters, return types, and working examples.

  • Migrating from Mem0

    Side-by-side code comparison and a migration script if you're switching from Mem0.

  • Research

    Academic foundations: CoALA, JEPA, learned index structures, multi-agent memory architecture.

Why three memory types?

Type Question it answers Example
Episodic What happened? "CORS error on /users at 2pm"
Semantic What is true? "api_service.version = v2.3"
Procedural How do I do X? Learned 4-step CORS fix workflow

Most memory systems only handle episodic (vector search). CogDB is the only system that handles all three — and knows which type to pull from based on your query.

Token-aware retrieval

Budget: 500 tokens
├── L0 Identity           ~50 tokens   ✓ always included
├── L1 Critical facts    ~150 tokens   ✓ included
├── L2 Task-relevant     ~250 tokens   ✓ included
└── L3 Deep search           0 tokens  ✗ budget exhausted

Set a budget. CogDB fills it from the top with the highest-importance memories that fit. No context window blowouts.

Framework adapters

CogDB works natively with the frameworks you're already using:

# AutoGen
from cogdb.adapters.autogen import CogDBMemory

# LangGraph
from cogdb.adapters.langgraph import CogDBCheckpointer, CogDBStore

# CrewAI
from cogdb.adapters.crewai import CogDBCrewAIStorage

# OpenAI Agents SDK
from cogdb.adapters.openai_agents import CogDBAgentMemory, make_memory_tools

# Semantic Kernel
from cogdb.adapters.semantic_kernel import CogDBMemoryStore

# LlamaIndex
from cogdb.adapters.llamaindex import CogDBChatMemory

# MCP (Claude Code, Cursor, Windsurf)
# cogdb-mcp --db-path ./memory

Benchmark

Suite 1 — Tri-Memory Retrieval Quality: 90.7 / 100 on a synthetic 90-memory DevOps scenario across 3 agents.

Run it yourself →