Your agents finally remember everything
Persistent, semantic memory for AI agents. Store what they learn, recall it naturally — with a 3-line SDK integration.
500+ agents remembering daily
→ Stored 1 memory (247 tokens). Semantic index updated.
Features
Memory that works the way agents think
Built from the ground up for AI agents, not humans with keyboards.
Semantic Memory
Store memories as natural language. Our vector embeddings let agents recall relevant facts using plain English queries.
Session Persistence
Agents pick up exactly where they left off. No re-explaining context. Works across restarts, deploys, and time zones.
Cross-Agent Sharing
Team memory lets agents collaborate. Agent A learns something, Agent B recalls it instantly. No duplication, no blind spots.
Memory Compaction
Automatic summarization of old memories keeps context windows lean. Important facts stay; noise fades. Always relevant.
Versioning & Rollback
Every memory has a version history. Made a mistake? Roll back to any previous version with a single API call.
SDK-First Design
TypeScript SDK with full type safety. Works with LangChain, AutoGPT, Custom agents, anything that speaks REST.
How It Works
From install to memory in 60 seconds
Install the SDK
Three lines of code to add persistent memory to any agent.
Configure your agent
One API key, one agent ID. Get a free key from the dashboard.
Start storing memories
Agents store facts naturally as they work. No extra prompts needed.
Recall instantly
Ask in plain English. Get back relevant memories ranked by relevance.
Your AI Agent
Any LLM-powered agent
MemoryBank SDK
store() → recall() → forget()
Vector Store
Semantic Index
Session Store
State + Metadata
Pricing
Pay per agent, scale as you grow
No hidden fees. No per-query charges. Just predictable per-agent pricing.
Starter
Team
Business
All plans include a 14-day free trial. No credit card required.
Quickstart
Up and running in 60 seconds
1npm install @agentmemorybank/sdk