I would love to see Remalt introduce My Second Brain: a persistent personal RAG knowledge layer that users can build once and use across Remalt, external AI tools, agents and workflows.
The idea is bigger than normal AI memory.
Instead of repeatedly uploading the same documents, explaining context or rebuilding prompts, Remalt could become the central knowledge layer containing what I know, research, save, write and approve.
1. Personal RAG Knowledge Base
Users could add:
• Documents and PDFs
• Notes and research
• URLs and web content
• YouTube/video transcripts
• Previous projects and conversations
• Writing samples and approved outputs
• Personal knowledge and experience
• Opinions, principles and established POV
• Decisions and reasoning
• SOPs, prompts and frameworks
• Examples and anti-examples
Remalt would index this content for semantic/hybrid retrieval and bring only the most relevant context into each task.
It should preserve source provenance so retrieved knowledge can show where it came from, rather than becoming an unexplained AI memory.
2. Chrome Extension for Knowledge Capture
The Remalt Chrome extension could become the capture layer for the Second Brain.
While browsing, I should be able to save:
• Full webpages
• Selected text
• Articles
• Research
• PDFs
• YouTube pages/transcripts
• Screenshots
• Quotes
• Notes added during capture
Ideally I could select:
Save to Second Brain → choose workspace/project → add tags/note → save
Remalt could then process, chunk, index and make that source retrievable through RAG.
This would turn everyday browsing and research into a continuous knowledge-building workflow rather than something I have to manually organise later.
3. MCP Access
The Second Brain should expose an authenticated MCP server.
This would allow MCP-compatible AI agents and tools to query my Remalt knowledge directly.
For example:
“Search my Second Brain for everything I have collected about AI-assisted writing.”
“Find my previous thinking about this product decision.”
“Retrieve the research and sources behind this argument.”
“Find relevant examples from my approved writing.”
MCP should ideally support granular permissions such as:
Read only
Search and retrieve knowledge.
Read/write
Allow authorised agents to create or update knowledge.
Sensitive actions should remain permission-controlled.
4. API Access
Please also expose the knowledge system through an API with user-generated API keys.
Useful endpoints could support:
• Add/update/delete sources
• Search
• Semantic retrieval
• Metadata filtering
• Retrieve source content
• Create knowledge records
• Manage collections/projects
• Retrieve citations/provenance
This would allow developers to use Remalt as the RAG layer behind their own applications, automations and AI agents.
API keys should support scopes, expiration/revocation and usage visibility.
5. Knowledge Governance
The Second Brain should not become an uncontrolled collection of AI-generated memory.
Users should be able to see and manage what Remalt knows.
Important controls:
• Source provenance
• Created/updated dates
• Project/workspace boundaries
• Tags and collections
• Private/shared status
• Approved vs unverified knowledge
• Delete/edit controls
• Duplicate detection
• Outdated-information warnings
• Conflicting information
• Source-level permissions
• AI-generated vs human-provided knowledge
Users should control what becomes persistent knowledge.
6. Continuous Learning Loop
The interesting part is connecting capture, work and learning:
Browse → Capture → Process → Index → Retrieve → Create → Review → Approve → Learn
For example, I could research a topic through Chrome and save useful sources.
Later I ask Remalt to write an article.
Remalt retrieves the relevant research, my previous writing, established POV, style guidance and approved examples.
I review the result.
Any genuinely useful new insight or approved writing could then be saved back to the Second Brain.
Over time, the knowledge base becomes more useful because it contains not only information I consumed, but also the thinking and work I approved.
Why I would use this
I do not want another AI memory feature that simply remembers random details from conversations.
I want a user-owned RAG layer for my digital knowledge.
Remalt could become the bridge between:
Chrome for capture → Second Brain for knowledge → RAG for retrieval → Remalt for creation → MCP/API for external AI agents and applications.
That would make Remalt much more than another AI workspace.
It would become the knowledge infrastructure underneath the tools I already use.
