Monitor usage and cost

Monitor Pinecone Assistant usage and cost, set monthly spend alerts, and track token usage across chat, context retrieval, and evaluation.

You can set up email alerts that track your organization’s monthly spending. An alert notifies the recipients you choose when spending reaches a threshold you set, and it resets at the start of each monthly billing cycle.

Spend alerts are available on the Standard and Enterprise plans. They’re not needed on the Starter or Builder plans, where usage is capped by plan quotas rather than billed per unit.

Set a spend alert

1. Go to Settings > Spend alerts in the Pinecone console.
2. Click + Add Alert.
3. Enter the dollar amount for the spend alert.
4. Enter the email addresses to send the alert to. Organization owners are listed by default.
5. Click Create.

Edit a spend alert

1. In the row of the spend alert you want to edit, click ellipsis (…) menu > Edit.
2. Change the dollar amount, the email addresses, or both.
3. Click Update.

Auto-spend spike alert

To protect against unexpected cost increases, Pinecone sends an alert when spending exceeds double your previous month’s invoice. You can’t change that threshold or delete the alert, but you can change which email addresses receive it and turn its notifications on or off.

Monitor organization-level usage

You must be an organization owner or billing admin to view usage across your Pinecone organization. Also, this feature is available only to organizations on the Standard or Enterprise plans.

Usage and costs across your organization are reported in the Usage dashboard in the Pinecone console. You can read the report there, download it as a CSV file, or export it with the Admin API (Enterprise plans only). For the steps, see Download a usage report.

Monitor token usage

Requests to the chat, context retrieval, and evaluation API endpoints return a usage parameter with prompt_tokens, completion_tokens, and total_tokens generated.

Chat

For chat, tokens are defined as follows:

• prompt_tokens are based on the messages sent to the assistant and the context snippets retrieved from the assistant and sent to a model. Messages sent to the assistant can include messages from the chat history in addition to the newest message.

  prompt_tokens appear as Assistants Input Tokens on invoices.

• completion_tokens are based on the answer from the model.

  completion_tokens appear as Assistants Output Tokens on invoices.

• total_tokens is the sum of prompt_tokens and completion_tokens.

Example chat response

{
    "finish_reason": "stop",
    "message": {
        "role": "assistant",
        "content": "The Chief Financial Officer (CFO) of Netflix is Spencer Neumann."
    },
    "id": "000000000000000030513193ccc52814",
    "model": "gpt-4o-2024-11-20",
    "usage": {
        "prompt_tokens": 23626,
        "completion_tokens": 21,
        "total_tokens": 23647
    },
    "citations": [
        {
            "position": 63,
            "references": [
                {
                    "file": {
                        "status": "Available",
                        "id": "99305805-3844-41b5-af56-ee693ab80527",
                        "name": "Netflix-10-K-01262024.pdf",
                        "size": 1073470,
                        "metadata": null,
                        "updated_on": "2025-07-29T20:07:53.171752661Z",
                        "created_on": "2025-07-29T20:07:36.361322699Z",
                        "signed_url": "https://storage.googleapis.com/..."
                    },
                    "pages": [
                        78,
                        79,
                        80
                    ],
                    "highlight": null
                },
                {
                    "file": {
                        "status": "Available",
                        "id": "99305805-3844-41b5-af56-ee693ab80527",
                        "name": "Netflix-10-K-01262024.pdf",
                        "size": 1073470,
                        "metadata": null,
                        "updated_on": "2025-07-29T20:07:53.171752661Z",
                        "created_on": "2025-07-29T20:07:36.361322699Z",
                        "signed_url": "https://storage.googleapis.com/..."
                    },
                    "pages": [
                        77,
                        78
                    ],
                    "highlight": null
                }
            ]
        }
    ]
}

Context retrieval

For context retrieval, tokens are defined as follows:

• prompt_tokens are based on the messages sent to the assistant and the context snippets retrieved from the assistant. Messages sent to the assistant can include messages from the chat history in addition to the newest message.

  prompt_tokens appear as Assistants Context Tokens Processed on invoices.

• completion_tokens don’t apply for context retrieval because, unlike for chat, there is no answer from a model. completion_tokens will always be 0.

• total_tokens is the sum of prompt_tokens and completion_tokens.

Example context response

{
    "snippets": [
        {
            "type": "text",
            "content": "edures, or caused such disclosure controls and procedures to be designed under our supervision, to\r\nensure that material information relating to the registrant, including its consolidated subsidiaries, ...",
            "score": 0.86632514,
            "reference": {
                "type": "pdf",
                "file": {
                    "status": "Available",
                    "id": "99305805-3844-41b5-af56-ee693ab80527",
                    "name": "Netflix-10-K-01262024.pdf",
                    "size": 1073470,
                    "metadata": null,
                    "updated_on": "2025-07-29T20:07:53.171752661Z",
                    "created_on": "2025-07-29T20:07:36.361322699Z",
                    "signed_url": "https://storage.googleapis.com/..."
                },
                "pages": [
                    78,
                    79,
                    80
                ]
            }
        },
        ...
    ],
    "usage": {
        "prompt_tokens": 22914,
        "completion_tokens": 0,
        "total_tokens": 22914
    },
    "id": "00000000000000007b6ad859184a31b3"
}

Response evaluation

For response evaluation, tokens are defined as follows:

• prompt_tokens are based on two requests to a model: The first request contains a question, answer, and ground truth answer, and the second request contains the same details plus generated facts returned by the model for the first request.

  prompt_tokens appear as Assistants Evaluation Tokens Processed on invoices.

• completion_tokens are based on two responses from a model: The first response contains generated facts, and the second response contains evaluation metrics.

  completion_tokens appear as Assistants Evaluation Tokens Out on invoices.

• total_tokens is the sum of prompt_tokens and completion_tokens.

Response evaluation response

{
  "metrics": {
    "correctness": 123,
    "completeness": 123,
    "alignment": 123
  },
  "reasoning": {
    "evaluated_facts": [
      {
        "fact": {
          "content": "<string>"
        },
        "entailment": "entailed"
      }
    ]
  },
  "usage": {
    "prompt_tokens": 123,
    "completion_tokens": 123,
    "total_tokens": 123
  }
}