Clay Icon

Concerns About Low Rate Limits for Claude 3 Opus Agent

·
·

Why are the rate limits for the clay managed anthropic account for their Claude 3 Opus agent so low? My AI columns in the below table keep erroring out and i have to manually re-run them because the rate limit keeps getting exceeded, which is taking up a lot of time. Is there a way to have the columns re-run themselves if errors exist? Or increase your rate limit? https://app.clay.com/workspaces/476916/workbooks/wb_FVwvdDHfoEr8/tables/t_CNjxUXsdUirP/views/gv_xDcsppHPkjgT

  • Avatar of Channeled
    Channeled
    APP
    ·
    ·

    The rate limits for the Clay managed Anthropic account are based on the API tier. With a lower tier, you'll need to process data in smaller batches, around 50 rows at a time, to avoid exceeding limits. To automatically re-run AI columns with errors: 1. Right-click the column and select "Play" 2. Choose "Run All Rows that haven't run or have errors" This will reprocess rows with errors without manual intervention. To manage your limits, you can: 1. Process data in smaller batches 2. Consider upgrading your Anthropic API tier for higher limits If you need further assistance, please request human support by clicking on "Talk to a person" in the chat.

    Source: Run Options

  • Avatar of Dave S.
    Dave S.
    ·
    ·

    That doesn’t solve my problem. Clay must be running an insane number of calls, I’m wondering why your rate limit is so low?

  • Avatar of Dave S.
    Dave S.
    ·
    ·

    I signed up for an anthropic API account, but I don’t have enough spend to justify getting to Tier 4 which seems to be what I would need.

  • Avatar of Bo (.
    Bo (.
    ·
    ·

    Hey, Totally understand how frustrating that can be. While the Clay-managed Anthropic account does have a high rate limit, it’s shared across many users—so hitting the cap can happen during peak usage. We’re actively working on increasing capacity, but you’re right: upgrading to a Tier 4 API on your own Anthropic account is currently the only way to get consistently higher throughput. As for re-running AI columns when errors occur—right now, it needs to be triggered manually, but I’ll share this feedback with the team. Let me know if you have more questions

  • Avatar of Dave S.
    Dave S.
    ·
    ·

    Ok, would really appreciate a solve here if possible soon!

  • Avatar of Bo (.
    Bo (.
    ·
    ·

    Hey, Totally understand. For now, the best option is to manually rerun the rows that encountered errors. This should help while we’re working on increasing the capacity to avoid these limits. Let me know if you need help walking through it!

  • Avatar of Channeled
    Channeled
    APP
    ·
    ·