🧠 MCP Qdrant Semantic Search

┌─────────────┐         ┌──────────────┐         ┌─────────────┐
│   Claude    │ ◄─MCP──►│  MCP Server  │ ◄─────► │   Qdrant    │
│   Desktop   │         │  (TypeScript)│         │  Vector DB  │
└─────────────┘         └──────────────┘         └─────────────┘
                              │
                              ▼
                        ┌──────────────┐
                        │    OpenAI    │
                        │  Embeddings  │
                        └──────────────┘

# 1. Install dependencies
npm install

# 2. Configure environment
cp .env.example .env
# Edit .env and add your OPENAI_API_KEY

# 3. Start Qdrant
docker-compose up -d

# 4. Build the project
npm run build

# 5. Configure Claude Desktop
# See INSTALL.md for details

Store this information: "Our API uses JWT for authentication.
Tokens expire after 24h and must be renewed via /refresh-token"

{
  "success": true,
  "message": "Memory stored successfully",
  "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "content": "Our API uses JWT for authentication..."
}

Search for how to handle user sessions

Store this code with tags "authentication" and "nodejs":

function validateToken(token) {
  try {
    return jwt.verify(token, process.env.JWT_SECRET);
  } catch (error) {
    throw new Error('Invalid token');
  }
}

Search for authentication code, only JavaScript snippets

Show me my semantic memory stats

{
  "success": true,
  "stats": {
    "name": "semantic_memory",
    "points_count": 42,
    "status": "green"
  },
  "embedding_model": "text-embedding-3-large",
  "embedding_dimensions": 1536
}

# Conceptual
"JWT authentication" → [0.234, -0.567, 0.891, ..., 0.123]
"Token security"     → [0.219, -0.543, 0.876, ..., 0.134]
# These two vectors are close = similar meaning!

{
  content: string,      // Content to store
  metadata?: {          // Optional metadata
    tags?: string[],
    category?: string,
    source?: string,
    // ... other fields
  }
}

{
  query: string,        // Natural language query
  limit?: number,       // Number of results (default: 5)
  threshold?: number,   // Min score 0-1 (default: 0.7)
  filter?: object       // Metadata filters
}

{
  id: string           // Memory ID
}

{
  id: string           // Memory ID
}

{
  limit?: number,      // Number of results (default: 10)
  offset?: string      // Starting ID for pagination
}

{
  confirm: boolean     // Must be true to confirm
}

Store these:

1. "Staging server accessible via staging.example.com,
    port 3000, credentials in 1Password"

2. "To deploy to production, use 'npm run deploy:prod'
    after tests pass and PR approval"

3. "Rate limiting is 1000 req/min per API key,
    10000/min for enterprise clients"

How do I deploy to production?
What are the API limits?

Store this code:

// Metadata: language=javascript, topic=authentication
async function authenticateUser(email, password) {
  const user = await db.users.findByEmail(email);
  if (!user) throw new Error('User not found');

  const valid = await bcrypt.compare(password, user.passwordHash);
  if (!valid) throw new Error('Invalid credentials');

  return generateJWT(user);
}

How to verify user credentials?
Show me login code

Store my preferences:
- I prefer TypeScript over JavaScript
- I use React 18 with hooks
- My code style follows Airbnb ESLint
- I want JSDoc comments on public functions

// In Claude
Search for authentication code,
only Python snippets created after 2024-01-01

const chunkSize = 500; // words
const chunks = splitIntoChunks(document, chunkSize);

for (const chunk of chunks) {
  await storeMemory({
    content: chunk,
    metadata: {
      document_id: "doc-123",
      chunk_index: i,
      total_chunks: chunks.length
    }
  });
}

# Check if Qdrant is running
docker ps | grep qdrant

# Restart if needed
docker-compose restart