Welcome to Hebbrix
Hebbrix gives your AI persistent memory. Build applications with durable memory, adaptive retrieval, and learning from reported outcomes.
01. Why Hebbrix?
Traditional LLMs forget everything after each conversation. Hebbrix fixes this by providing a memory layer that stores, retrieves, and can learn from reported outcomes. Your AI can now remember user preferences, past conversations, and build knowledge temporal graphs that preserve how facts and relationships change over time.
Versioned Memory
Operational tiers, current truth, history, provenance, and explicit deletion controls.
Smart Retrieval
Hybrid search finds exactly what you need, even from old conversations.
Drop-in Ready
Works with OpenAI, Anthropic, and any LLM. Just change the base URL.
02. Explore the API
Memory Management
Store, retrieve, version, and manage current and historical AI memory.
Hybrid Search
Adaptive semantic, lexical, graph, temporal, and calibrated retrieval.
Chat Completions
OpenAI-compatible chat API with automatic memory injection.
Collections
Organize memories into collections for multi-tenant applications.
Knowledge Graph
Automatic entity extraction and relationship mapping.
Outcome Memory
Decision receipts, delayed outcomes, posterior uncertainty, and conservative evaluation.
MCP Server
Connect any MCP-compatible agent through one hosted endpoint with 26 live tools.
Coding Agent
MCP integration for Claude, Cline, and any MCP-compatible agent.
Privacy Controls
GDPR-aligned data export, deletion, and access-management controls.
03. Quick Example
import asyncio
from hebbrix import MemoryClient
async def main():
async with MemoryClient(api_key="mem_sk_...") as client:
coll = await client.collections.create(name="my-agent")
# Store a memory
await client.memories.create(
collection_id=coll["id"],
content="User prefers dark mode",
)
# Search memories
results = await client.search(
query="user preferences",
collection_id=coll["id"],
)
asyncio.run(main())04. Ready to Build?
Get started in minutes with our quickstart guide or explore the full API reference.