Tidemind
About
Open-source AI memory layer — a living knowledge graph across all your AI tools and notes. Local-first, SQLite-backed, MCP-native.
Details
- License
- MIT
Explore
- All memory in a single knowledge graph.
- Profile built from all data delivered to AI at conversation start.
- Graph-based recall that follows links between memories.
- Background metabolism that reinforces, crystallizes, and forgets memories.
- Local SQLite storage, exportable to Markdown anytime.
- Integrates with multiple AI tools and note systems.
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
TidemindCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
- Node.js >= 18
That's the only hard requirement. TideMind runs out of the box with text search (BM25) and basic storage.
Optional, for the full experience:
- An LLM provider (Anthropic API Key, Google Vertex AI, or Gemini) — enables automatic information extraction, memory reconsolidation, divergent scanning, and crystal emergence. Without it, these features are simply skipped.
- An embedding provider (Ollama, Vertex AI, or Gemini) — enables semantic vector search alongside BM25. Without it, search falls back to keyword matching.
- You can mix and match: e.g., Ollama locally for embeddings + Anthropic for LLM, or go fully local, or fully cloud. See the Integration Guide for details.
brain_prepare
Load memory context
brain_recall
Navigate and retrieve memories
brain_digest
Store new information
Cursor
Supported
Windsurf
Supported
Codex
Supported
TideMind exposes three tools via the MCP protocol, corresponding to three cognitive operations:
| Tool | What it does | When to use |
|------|-------------|-------------|
| brain_prepare | Load memory context | Start of every conversation |
| brain_recall | Navigate and retrieve memories | When you need historical context |
| brain_digest | Store new information | When something worth remembering emerges |
This isn't CRUD. Prepare is like opening a memory map. Recall is like remembering — and each recall strengthens the memory. Digest is like absorption — incoming information is decomposed, connected, and woven into the existing knowledge network.
brain_recall goes beyond search. AI can follow links between memories — starting from an architecture decision, jumping to the discussion behind it, then to later changes it triggered. Like walking through a knowledge network, not guessing keywords in a search box. It returns a small set of highly relevant nodes first, then expands along links when more context is needed — exploring on demand, never loading everything upfront.
Learn more about the three tools
| Tool | Status | Notes |
|------|--------|-------|
| Claude Code | Supported | Native MCP support |
| Cursor | Supported | MCP configuration |
| Windsurf | Supported | MCP configuration |
| Codex | Supported | MCP configuration |
| Any MCP-compatible tool | Supported | Standard protocol, plug and play |
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"tidemind": {
"tidemind": {
"command": "node",
"args": [
"/path/to/tidemind/dist/index.js"
],
"env": {
"EB_AGENT_ID": "claude-code-main"
}
}
}
}
}McpServers
{
"tidemind": {
"command": "node",
"args": [
"/path/to/tidemind/dist/index.js"
],
"env": {
"EB_AGENT_ID": "claude-code-main"
}
}
}Why TideMind
Your memory is scattered across a dozen tools, and none of them talk to each other.
Claude remembers you like concise code, but Cursor has no idea. You spent a week organizing design thinking in your notes, but the next time you discuss it with an AI, it knows nothing. Every tool quietly accumulates its own understanding of you, yet nothing connects them.
The deeper problem: existing AI memory is just listing and storing facts. It won't link memories together, won't discover commonalities across your ideas that you haven't noticed yourself, and certainly won't build a complete profile of who you are. And as memories pile up, the data grows increasingly redundant and harder to retrieve.
TideMind is a memory layer that spans all your tools — a living knowledge graph that connects your AI, your notes, and your thinking.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
