Nowledge Mem
Nowledge Mem helps sink, exchange and augment your personal context among different tool, agents Purely computed and storaged locally with only on-device AI models.
Read more via the blog here.
Creating Memories
There are three main ways to create a new memory:
1. Import from AI Agent Conversations
This is the most common way to create memories. After configuring MCP (Model Context Protocol) with your AI agent, you can easily extract insights from your conversations.
How to do it:
- During your conversation with an AI agent, simply tell it:
- "Summarize our recent insight and add it to memory"
- "Summarize our recent decision and add it to memory"
- "Save this key finding to Nowledge Mem"
- The agent will use MCP to automatically create a memory in Nowledge Mem
Like this:
This clip shows how to create a memory from a conversation with Cursor, and then consume the memory in a new conversation.
Read more: Integrations: MCP
2. Distill Memories from Conversation Threads
When you have long, valuable conversation threads from ChatGPT, Gemini, Cursor, Claude, or other AI tools, you can distill multiple insights into memories in batch.
Two ways to import threads:
From Exported Files
- Export your conversation thread from your AI tool
- Import the file into Nowledge Mem
- Select which parts to distill into memories
Browser Extension (One-Click Import)
- Install the Nowledge Mem browser extension
- Navigate to your conversation thread webpage
- Click the extension to import directly
3. Create Memories Manually
Navigate to Memories view, and click the "+ Create" button to create a new memory.
What Happens After Creating a Memory?
Once you create a memory from your agent, Nowledge Mem automatically indexes it for fast search (BM25 + vector search).
If needed, you can also augment the memory later with knowledge graph, which:
-
enhanced the searching of memories leveraging the connections and global view of knowledge among memories
-
allow us to explore your memories within the knowledge graph
If we instead import the thread directly, and distill it into memories, there is an option to augment the memory with knowledge graph together with the memory distillation.
Next Steps
Now that you know how to create memories, learn how to find and use them effectively? See more: https://mem.nowledge.co/docs/usage
Server Config
{
"mcpServers": {
"nowledge-mem": {
"url": "http://localhost:14242/mcp",
"type": "streamableHttp",
"headers": {
"APP": "Client via MCP.so"
}
}
}
}