Skip to main content
Docs
Feature Requests Onboarding Trust Center Status Page

Community

Anthropic Claude Connector — Missing columns on cost_report_result and message_usage_report_result

Completed
Maria Pradeep Irudayaraj User

Request 1:

Connector name: Anthropic Claude (Lite, Private Preview)

Table name: MESSAGE_USAGE_REPORT_RESULT

Columns to add: cache_creation_ephemeral_1h_input_tokens, cache_creation_ephemeral_5m_input_tokens

API documentation link: https://docs.anthropic.com/en/api/admin-api/usage-cost/get-messages-usage-report

Additional details: The Anthropic Messages Usage Report API returns a cache_creation object with ephemeral_1h_input_tokens and ephemeral_5m_input_tokens fields in each result. These are not currently synced to the MESSAGE_USAGE_REPORT_RESULT table. We need these to accurately measure cache utilization and estimate cost savings from prompt caching. The table currently has cache_read_input_tokens and uncached_input_tokens but is missing the cache write/creation side



Request 2:

Connector name: Anthropic Claude (Lite, Private Preview)

Table name: COST_REPORT_RESULT

Columns to add: model, token_type, cost_type, context_window, inference_geo

API documentation link: https://docs.anthropic.com/en/api/admin-api/usage-cost/get-cost-report (not publicly documented yet, but the CSV export from Anthropic console includes these columns)

Additional details: The Anthropic Cost Report CSV export includes granular breakdowns by model (e.g., Claude Haiku 4.5, Claude Opus 4), token_type (input_no_cache, input_cache_read, input_cache_write_5m, output), and cost_type (token, web_search). The COST_REPORT_RESULT table currently only has amount and currency without these dimensions, making it impossible to attribute costs by model or token type. This is critical for cost management across multiple Claude models with different pricing.

Please sign in to leave a comment.

Comments

3 comments

Date Votes
  • Official comment
    Sisan Kumar Sahu Fivetranner

    Hi Maria,

    Hope you're doing well.

    I'm happy to let you know that all the concerns from the incidents you raised have been resolved, and I can see you've acknowledged them in the support ticket. Thank you for confirming.

    Regarding the field you mentioned: `cache_creation_input_token` isn't a supported field in Claude at this time. I believe you may have been referring to `cache_creation_ephemeral_1h_input_tokens` and `cache_creation_ephemeral_5m_input_tokens` (or `cache_creation_input_tokens`), which should already be populated for you by now.

    We truly appreciate your feedback and continued partnership - it plays an important role in shaping our product roadmap.

    Best regards,
    Sisan

  • Sisan Kumar Sahu Fivetranner
    • Edited

    Hello Maria,

    Thank you for raising the request regarding the Anthropic Connector.

    Request 1:
    Table: MESSAGE_USAGE_REPORT_RESULT

    The fields cache_creation_ephemeral_1h_input_tokens and cache_creation_ephemeral_5m_input_tokens are currently available under the names:

    • results_cache_creation_ephemeral_1h_input_tokens
    • results_cache_creation_ephemeral_5m_input_tokens

    Could you please verify from your end whether these fields are being populated?

    Request 2:
    We are in the process of updating the connector to include the requested fields. We will share an update as soon as the fix is deployed.

    We appreciate your feedback and continued support - it plays an important role in shaping our product roadmap.

    Best regards,
    Sahu

  • Maria Pradeep Irudayaraj User

    Hi,

    Thanks for the update.

    Regarding Request 1 -  I checked both results_cache_creation_ephemeral_1h_input_tokens and results_cache_creation_ephemeral_5m_input_tokens on the Snowflake side. These columns do not exist on the MESSAGE_USAGE_REPORT_RESULT table. I also ran a full resync of both MESSAGE_USAGE_REPORT and MESSAGE_USAGE_REPORT_RESULT and the columns still did not appear.

    Additionally, the following fields from the API schema are also not populating in Snowflake but showing up on Fivetran schema:

    • MESSAGE_USAGE_REPORT_RESULT: speed, inference_geo, cache_creation_input_token




      Could you please check this please?