Pinecone MCP Server — Hosted for Any AI Agent

mcp-pinecone — Pinecone MCP Server

Query, upsert, and manage vectors in your Pinecone indexes from your AI agents.

Pinecone is the leading vector database for AI/ML applications — purpose-built for high-performance similarity search at scale. This MCP server gives your AI agents full access to your Pinecone indexes: querying vectors for semantic search and RAG, upserting embeddings, managing namespaces, and inspecting index statistics — all through natural language.

Live endpoint: https://mcp.aerostack.dev/s/aerostack/mcp-pinecone

---

What You Can Do

• Query vectors by embedding or ID for semantic search, recommendation, and RAG retrieval workflows
• Upsert embeddings with metadata to build and maintain your vector knowledge base
• Manage indexes — list, describe, and inspect statistics across all your Pinecone indexes
• Delete vectors by ID, metadata filter, or wipe entire namespaces for data lifecycle management
• Paginate through vector IDs and fetch full vectors by ID for auditing and debugging

Available Tools

Tool	Description
list_indexes	List all indexes with status, dimension, metric, and host
describe_index	Get detailed info for a specific index (host, dimension, status)
query	Query vectors by embedding or ID with filters, topK, and metadata
upsert	Upsert vectors with IDs, values, and optional metadata
fetch	Fetch full vectors by their IDs
delete_vectors	Delete vectors by IDs, metadata filter, or delete all
describe_stats	Get index statistics — total vectors, per-namespace counts
list_vectors	List vector IDs in a namespace with pagination
update_vector	Update a vector's values or metadata by ID

Configuration

Variable	Required	Description	How to Get
PINECONE_API_KEY	Yes	Your Pinecone API key	app.pinecone.io → API Keys → copy your key

Quick Start

Add to Aerostack Workspace

1. Go to aerostack.dev → Your Project → MCPs
2. Search for "Pinecone" and click Add to Workspace
3. Add PINECONE_API_KEY under Project → Secrets

Once added, every AI agent in your workspace can call Pinecone tools automatically — no per-user setup needed.

Example Prompts

"List all my Pinecone indexes and show their dimensions and status"
"Query my product-embeddings index for the top 5 vectors similar to this text embedding"
"Upsert these 3 document embeddings into the knowledge-base index with their source metadata"
"Show me the stats for my main index — how many vectors per namespace?"

Direct API Call

curl -X POST https://mcp.aerostack.dev/s/aerostack/mcp-pinecone \
  -H 'Content-Type: application/json' \
  -H 'X-Mcp-Secret-PINECONE-API-KEY: your-pinecone-api-key' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"list_indexes","arguments":{}}}'

License

MIT