Every conversation starts fresh. Your AI forgets your name, your preferences, everything you've ever told it. Engrm fixes that.
Session 1
"I'm John, based in Singapore. I prefer morning meetings."
Session 2
"Schedule a call with my team."
"What timezone are you in? Do you have any scheduling preferences?"
Session 3
"Remember that project we discussed?"
"I don't have access to previous conversations."
Every session starts from zero.
Session 1
"I'm John, based in Singapore. I prefer morning meetings."
✓ Stored: identity, preference, timezone
Session 2
"Schedule a call with my team."
"I'll suggest 9am SGT — that works with your morning preference and gives US teammates a reasonable evening slot."
Session 47
"What did we decide about the API?"
"You chose REST over GraphQL on Feb 15th because of team familiarity."
Context builds over time. Nothing is forgotten.
Every conversation, Engrm retrieves relevant context and injects it into your AI's prompt. Your agent responds like it actually knows you.
User Message
User asks a question
Search Memory
Find relevant context
Decrypt Local
Only you can read it
Inject Context
Add to AI prompt
Smart Response
Personalized answer
System:
You are a helpful assistant.
## Memories (auto-injected by Engrm):
User:
"Schedule a call with the team"
The AI now has full context to give a personalized, relevant response.
Engrm doesn't just store memories — it learns what matters, forgets what doesn't, and builds connections between related ideas.
Mention something multiple times? It becomes a stronger memory. Important things naturally rise to the top.
Not everything is worth keeping forever. Old, unused memories naturally fade — just like human memory.
Encrypted on your device. We literally cannot read your memories — even with full database access.
Works with Claude, GPT, or any LLM. MCP server, Python SDK, or REST API.
For Claude Desktop & Cursor
npm install -g @engrm/mcp
// claude_desktop_config.json
"engrm": {
"command": "engrm-mcp"
}Full ZK with local embeddings
pip install engrm-sdk
client = MemryClient(
api_key="mem_xxx",
vault_password="***"
)
client.store("User prefers dark mode")Any platform, any language
POST /api/v1/memories
{
"content": "encrypted...",
"embedding": [0.1, ...],
"type": "preference"
}
POST /api/v1/context
{ "query_embedding": [...] }Free to start. No credit card required. Your data stays encrypted.