Agents don't know what they should know.
When building agents, knowledge ends up scattered.
Each agent ends up with its own partial view of the world.
There's no shared structured source of truth.
Memory engine fixes this:
curl -sSfL https://install.memory.build | sh && me installMemory Engine
is the shared memory layer for your agents.
Store context once. Retrieve it precisely when it's needed. Agents, tools, and developers can all access the same memory — with control over who sees what.
Your agents stop guessing. They retrieve.
Find context the way agents actually need it.
Memory Engine supports multiple retrieval methods that can be used independently or combined.
Agents — or developers — decide what retrieval strategy works best.
An MCP server for all your tools
TypeScript client available for agents and applications.
Works locally
Memory Engine runs a local embedding model by default. nomic-ai/nomic-embed-text-v1.5 via ONNX, in-process.
Built directly on PostgreSQL
Deploy how you want
SOLO: install and go. Runs locally using a ghost database.
TEAM: run a central server and connect agents and users via API keys.
Multi-agent.
Multi-user.
Agents, tools, and developers can all access the same memory.
Access is controlled through scoped permissions.
Agents see only the context they're allowed to see.
Instant knowledge for your agents.
Memory Packs are pre-built bundles of knowledge and skills.
Import a pack and your agents immediately gain structured memories they can retrieve.
Get started
Paste onto your terminal to install Memory Engine