bluecolumn-memory
LocalAbout bluecolumn-memory
bluecolumn-memory is an MCP server that gives AI agents persistent semantic memory across sessions, including text, documents, audio, and music. Its tools support natural-language recall, audio analysis, and music-specific search with structured musical context.
Key features
- 01Persistent memory across sessions
- 02Ingest text, audio, and documents
- 03Natural-language recall with sources
- 04Audio analysis and event indexing
- 05Music memory with musical filters
Use cases
- 01Recall project context in a new session
- 02Store and search voice notes or meetings
- 03Track practice sessions and musical techniques
MCP support
Exposes memory, audio, and music tools to MCP-compatible clients. The tools include storing observations, recalling memories, preparing for and recording calls, and searching streamed audio.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y bluecolumn-mcpBLUECOLUMN_API_KEYSecretYour BlueColumn API key. Get a free key at https://bluecolumn.aiBLUECOLUMN_API_URLBlueColumn API base URL for the audio-intelligence layer (/v1/audio/*, /v1/calls/*, /v1/sound/*, /v1/music/*). Defaults to http://localhost:8000 (local FastAPI backend).Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
14Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| text | string | No | Raw text content to store in memory |
| title | string | No | Optional title for this memory (include date for best recall) |
| file_url | string | No | URL to a PDF or document |
| audio_url | string | No | URL to an audio file (will be transcribed via Whisper) |
Details
From the project README🧠 BlueColumn MCP Server
Give Claude Desktop, Cursor, and any MCP-compatible agent persistent memory across sessions.
2-minute setup. Zero-config. Audio-native. Works with every MCP client.
MCP server for BlueColumn — persistent semantic memory for AI agents.
Give any MCP-compatible agent (Claude Desktop, Cursor, Windsurf, LangChain, AutoGen, CrewAI) the ability to remember, recall, and store observations across sessions.
Tools
Core memory
| Tool | Description |
|---|---|
remember | Ingest text, audio, or documents into persistent memory |
recall | Query memory with natural language, get AI-synthesized answer + sources |
note | Store lightweight agent observations as searchable vectors |
namespace | Show the configured namespace |
Music Memory (new)
| Tool | Description |
|---|---|
music_remember | Store a musical recording, practice session, or lesson with rich musical context — instrument, key, tempo, technique tags, chord progression, notes. Audio is transcribed; the structured context is embedded alongside so a coach/teacher agent recalls it with full musical understanding |
music_recall | Search stored musical content with musical filters — instrument, technique, or style. E.g. "show me every take with barre chord issues" or "what did we practice in E minor last month" |
Audio Intelligence
| Tool | Description |
|---|---|
audio_ingest | Ingest calls, voice notes, podcasts, meetings, or music and extract semantic memory |
audio_recall | Search audio memories and get audio-backed citations |
call_prepare | Get memory context before a voice call starts |
call_complete | Store what changed after a call ends |
sound_analyze | Detect and index non-speech audio events |
music_analyze | Extract tempo, structure, instrumentation, mood from music |
Streaming audio memory
| Tool | Description |
|---|---|
streaming_audio_ingest | Ingest edge-device audio chunks (car, doorbell, wearable) into per-device streaming memory |
streaming_audio_recall | Recall over a device's streamed audio history |
Music Memory example
Store a practice session:
music_remember {
"audio_url": "https://example.com/take3.mp3",
"title": "Week 3 practice — barre chords, Sep 7",
"instrument": "guitar",
"musical_key": "E minor",
"tempo_bpm": 92,
"technique_tags": ["barre-chords", "hammer-on"],
"style_tags": ["blues"],
"chord_progression": ["Em", "Am", "B7"],
"notes": "F-shape barre still buzzy on the B string; Fmaj7 subst worked well"
}
Then recall it musically:
music_recall { "q": "which takes had barre chord issues?", "instrument": "guitar" }
Setup
1. Get your BlueColumn API key
Sign up free at bluecolumn.ai — 60 min audio + 100 queries/month, no credit card required.
2. Install
npm install -g bluecolumn-mcp
3. Configure Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"bluecolumn": {
"command": "bluecolumn-mcp",
"env": {
"BLUECOLUMN_API_KEY": "bc_live_YOUR_KEY"
}
}
}
}
4. Restart Claude
Done. Claude now remembers everything you tell it to.
What Makes BlueColumn Different
| Feature | BlueColumn MCP | Other MCP Memory Servers |
|---|---|---|
| Audio ingestion | ✅ Native (transcribe + store) | ❌ Text only |
| Plain English recall | ✅ "What did we decide about pricing?" | ❌ Keyword/vector only |
| Setup time | ~2 minutes | ~10-30 minutes |
| Cloud managed | ✅ No infra to run | ⚠️ Most require local DB |
| Namespace isolation | ✅ Built-in per key | ⚠️ Manual |
Examples
Store project context
Remember: We're building a React component library called Aether with Tailwind CSS v4.
Recall across sessions (new conversation)
What project was I working on? What was my tech stack?
→ Claude recalls: "You're building Aether, a React component library with Tailwind CSS v4"
Store audio notes
Record a voice memo → BlueColumn transcribes it → Claude remembers it.
Advanced
Custom namespace
{
"mcpServers": {
"bluecolumn": {
"command": "bluecolumn-mcp",
"env": {
"BLUECOLUMN_API_KEY": "bc_live_YOUR_KEY",
"BLUECOLUMN_NAMESPACE": "my_project"
}
}
}
}
Multiple environments
Use different namespaces for work, personal, and side projects. Each stays isolated.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.