Last updated 07-30-2026
Category:
Reviews:
Join thousands of AI enthusiasts in the World of AI!
Memori Labs (formerly GibsonAI)
Memori Labs builds agent-native memory infrastructure that captures each agent run and chat turn as structured facts, preferences, rules, and summaries inside SQL databases you already operate. Drop the Memori SDK into existing agent code and the layer classifies what to remember, recalls only relevant context, and explains why each memory surfaced in a prompt.
Vector-only memory stacks replay long chat histories or bolt on separate search services. Memori stores memory in relational tables, runs targeted recall across conversations and documents, and adds semantic search only when phrasing gets fuzzy. The team reports 87 percent accuracy on LoCoMo with 97 percent fewer tokens than full-context retrieval, which matters when production agents run thousands of sessions per day.
The open-source Memori engine on GitHub pairs with Memori Cloud for teams that want a hosted API without wiring storage themselves. Production plans start at $60,000 per year for a single agent, with enterprise contracts for multi-agent fleets, SLAs, and forward-deployed engineering.
Open-source Memori SDK on GitHub with one-line enable() integration for existing LLM calls
LoCoMo benchmark shows 87 percent accuracy with 97 percent lower token use than full-context retrieval
Classifies each turn into facts, preferences, rules, and summaries with configurable retention
Memori Cloud free tier at $0 for prototyping with signup at app.memorilabs.ai
Production plans start at $60,000 per year for a single production agent
Multi-agent production tier starts at $150,000 per year for several live agents
Explainable recall includes lineage showing why each memory entry was included
SQL-native storage keeps agent memory portable, auditable, and queryable like any app data
Open-source Memori engine on GitHub enables self-hosting before committing to cloud pricing
Targeted recall and explainable lineage reduce token burn versus dumping full chat history
LLM-agnostic proxy integrates without replacing existing model providers or frameworks
Production pricing starts at $60,000 per year, which is steep for early-stage teams
gibsonai.com redirects to memorilabs.ai, which can confuse users following the old domain
Screenshot and marketing assets now use Memori branding while the listing slug remains gibsonai
What is Memori Labs?
Memori Labs, formerly GibsonAI, provides agent-native memory infrastructure called Memori. The layer turns agent execution and conversation into structured, persistent state inside SQL databases while staying LLM-agnostic across models and frameworks.
Is Memori free to try?
Yes, Memori Labs offers a Cloud Free plan at $0 for prototyping through app.memorilabs.ai. The open-source Memori SDK also lets teams bring their own database before moving to paid production tiers.
How much does Memori production cost?
Memori Labs lists production pricing from $60,000 per year for a single production agent and from $150,000 per year for multiple production agents. Enterprise contracts add custom SLAs, advanced evaluation, and forward-deployed engineering.
What databases does Memori support?
Memori Labs stores memory in standard SQL databases such as PostgreSQL, MySQL, and SQLite through the open-source engine. Teams can self-host the SDK or use Memori Cloud when they prefer a managed API.
How does Memori reduce token costs?
Memori Labs pulls only relevant memories instead of replaying entire chat histories. Benchmarks on LoCoMo report 87 percent accuracy while using 97 percent fewer tokens than full-context retrieval approaches.
Does Memori work with any LLM?
Yes, Memori Labs is LLM-agnostic and intercepts queries across models through a drop-in proxy or SDK hooks. The same memory graph can serve multiple agents and sessions without replacing your existing inference stack.