Skip to main content

​
At a glance

  • Availability: Experimental (how to enable).
  • Authentication: API key.
  • Connection: The key from PINECONE_API_KEY is sent as the Api-Key header.
  • Provider documentation: Authentication reference.

​
Credentials

Set these per environment. See Connect an integration.

​
Setup

  1. Create a Pinecone account: Sign up at https://app.pinecone.io. The free Starter plan includes serverless indexes for testing.
  2. Create an API key: In the Pinecone console, select your project, open API Keys, and create a key (pcsk_…). Copy it immediately.
  3. Store the key: Add it to your .env file as PINECONE_API_KEY=pcsk_…
  4. Verify access: Run the List Indexes tool. To read or write vectors, first run Describe Index for the index you want, take the returned ‘host’ value, and pass it without the trailing ‘.pinecone.io’ as the indexHostPrefix parameter of Query Vectors / Upsert Vectors.

​
Provider notes

  • Pinecone is two-step: control-plane tools (list/describe/create index) hit api.pinecone.io, while Query Vectors and Upsert Vectors target the index named by their indexHostPrefix parameter; each call may address a different index, and requests can only reach ‘<indexHostPrefix>.pinecone.io’ hosts, never an arbitrary domain
  • Migrating from earlier versions: the old indexHost parameter (full host) and the interim PINECONE_INDEX_HOST environment variable are gone. Pass the Describe Index host without its ‘.pinecone.io’ suffix as indexHostPrefix instead
  • All requests send the X-Pinecone-API-Version header (default 2025-01)
  • Embeddings must be generated separately (e.g. with OpenAI); vectors you query or upsert must match the index dimension
Provider API reference.

​
Tools

​
Verify the connection

Call a read tool such as pinecone__list_indexes with arguments for your account. Confirm that the result comes from the intended account or workspace before enabling write tools.