Troubleshooting

Fix API Rate Limits and Billing Errors on Your ProductiveBot

Your ProductiveBot stopped responding? Fix Claude Opus 4.6 API errors: check Anthropic credits, resolve 401/402/429/529 error codes, understand rate limit tiers, and explore alternative models. Step-by-step troubleshooting guide with cost estimates.

PB
ProductiveBot Team· · 18 min read
Troubleshooting Guide

Your ProductiveBot stopped responding, and you are seeing an error message about rate limits or API issues. This guide walks you through every common cause, how to check each one in Anthropic's dashboard, and what to do about it.

ProductiveBot runs on Claude Opus 4.6, Anthropic's most capable model. Opus 4.6 offers a 1-million-token context window, advanced reasoning, and strong performance on complex tasks. But like any API-powered service, it depends on your Anthropic account being in good standing.

Most of the time, when your bot stops working, it is not a ProductiveBot problem. It is an issue between your Anthropic API account and Anthropic's servers. The good news is that you can usually fix it yourself in a few minutes.

Quick check before you start: Open Terminal on your Mac Mini and run openclaw gateway logs. Look for the error code number (like 401, 402, 429, or 529). Then jump to that section below.

Understanding the Error Codes

When your bot stops working, the underlying error usually falls into one of these categories. Each one has a different cause and a different fix. These codes come directly from Anthropic's API error documentation.

401 — Authentication Error

Your API key is invalid or missing

This means Anthropic does not recognize the API key your ProductiveBot is using. The key may have been deleted, rotated, or never set up correctly.

  • CommonThe API key was deleted from your Anthropic account (sometimes by accident when cleaning up old keys).
  • SometimesThe key was never saved properly during initial setup. The setup wizard completed, but the key did not persist to the config file.
  • RareAn OpenClaw update overwrote the config file where the key was stored.
402 — Billing Error

Your credits ran out or payment failed

Anthropic's API is prepaid. You buy credits, and every message your bot sends uses some of those credits. When the balance hits zero, everything stops. Claude Opus 4.6 costs $5 per million input tokens and $25 per million output tokens (see current pricing), so credits can deplete faster than you might expect with heavy use.

  • CommonYour credit balance is $0. This is the most frequent cause of "my bot stopped working."
  • SometimesYour credits expired. Anthropic API credits expire after one year from purchase. Many people do not realize this.
  • SometimesA payment method failed or was removed from your Anthropic account.
429 — Rate Limit Error

Too many requests in a short time

Your bot sent more messages than your account tier allows within a one-minute window. Opus 4.6 shares rate limits with other Opus models (4.5, 4.1, and 4), so the limits apply to combined traffic across all Opus versions.

  • CommonYou are on Tier 1, which only allows 50 requests per minute and 30,000 input tokens per minute for Opus models. A single bot conversation with tools (web search, memory, file reading) can generate 5 to 10 API requests behind the scenes.
  • SometimesAnthropic's "acceleration limits" triggered because your usage ramped up too quickly from near zero. Even if you are well under your tier limit, a sudden burst can trigger a 429.
  • RareYou hit your monthly spend limit, which acts like a cap even if you still have credit remaining.
529 — Overloaded

Anthropic's servers are too busy

This one is not your fault at all. Anthropic's infrastructure is temporarily overloaded. This happens during peak usage times and usually resolves within 2 to 5 minutes. Opus 4.6 can be more prone to this than smaller models because it requires more compute per request.

  • CommonHigh global demand on Claude, especially on weekday afternoons (US time zones) and after new model launches.
  • RareA broader outage at Anthropic. Check status.anthropic.com if it lasts more than 10 minutes.

Step-by-Step: Check Your Anthropic Account

Before trying anything else, log into Anthropic's console and check these three things in order. This resolves the issue about 90% of the time.

1

Check Your Credit Balance

Go to console.anthropic.com/settings/billing and sign in with the same email you used when setting up your ProductiveBot.

Look at your credit balance. If it says $0.00 or shows a negative balance, that is why your bot stopped working.

How to add credits

  1. On the Billing page, click "Add Credits" (or "Buy Credits").
  2. Enter an amount. For Opus 4.6 usage, we recommend at least $25 to $50 to start. This typically lasts 2 to 4 weeks of normal use.
  3. Complete the payment with a credit or debit card.
  4. Your bot should start responding again within a minute or two. No restart needed.
API credits expire after one year. If you loaded credits onto your account a year ago, that balance may have expired even if you did not use it all. Anthropic does not always make this obvious in the dashboard. Check the "Credit Expiry" section on your billing page, or look for an email from Anthropic about expiring credits.
2

Check Your API Key

Go to console.anthropic.com/settings/keys.

You should see at least one API key listed. If the list is empty, your key was deleted and you need to create a new one.

If your key is missing or you need a new one

  1. Click "Create Key" on the API Keys page.
  2. Give it a name like "ProductiveBot" so you remember what it is for.
  3. Copy the key immediately. You will only see it once.
  4. On your Mac Mini, open Terminal and paste these two commands:
openclaw config set providers.anthropic.apiKey YOUR_NEW_KEY_HERE
openclaw gateway restart

Your bot should start responding again right away. The openclaw config set command saves the key to your local configuration, and the gateway restart picks up the change.

3

Check Your Usage Tier and Limits

Go to console.anthropic.com/settings/limits.

This page shows two important things:

  • Your usage tier (Tier 1 through Tier 4). Higher tiers allow more requests per minute.
  • Your monthly spend limit. This is the maximum Anthropic will let you spend in a calendar month, even if you have credits remaining.

Anthropic's usage tiers (Opus 4.6 limits)

Tier To qualify Monthly limit Requests/min Input tokens/min
Tier 1 $5 purchased $100/mo 50 30,000
Tier 2 $40 purchased $500/mo 1,000 450,000
Tier 3 $200 purchased $1,000/mo 2,000 800,000
Tier 4 $400 purchased $200,000/mo 4,000 2,000,000

Opus rate limits are shared across Opus 4.6, 4.5, 4.1, and 4. "To qualify" shows cumulative credit purchases needed.

These limits come from Anthropic's rate limits documentation and are current as of March 2026. Most ProductiveBot customers start at Tier 1. The 50 requests per minute limit sounds like a lot, but your bot does more behind the scenes than you might think. A single user message can trigger 5 to 10 API calls: the main response, plus tool calls for memory search, web lookups, or file reading. In a busy group chat, this adds up fast.

The good news: your tier increases automatically as you purchase more credits. You do not need to apply. Buying $40 total in credits moves you to Tier 2 (1,000 requests/minute), which is more than enough for almost any ProductiveBot use case.

Check your custom spend limit too. On the Limits page, Anthropic lets you set a monthly spending cap lower than your tier allows. If you set it to $10/month and your bot has been busy, you may have hit your own cap even with credits remaining. Raise it or remove it to restore access.

What If It Is a 529 (Overloaded)?

If your gateway logs show a 529 error, there is nothing wrong with your account. Anthropic's servers are temporarily overloaded.

Opus 4.6 requires significantly more compute than smaller models like Sonnet or Haiku. During peak demand, Anthropic may have capacity for lighter models but not enough for Opus. This means your bot might hit 529 errors while other Claude-powered tools (using Sonnet) keep working normally.

What to do

  • Wait 2 to 5 minutes. Most 529 errors resolve on their own. Your bot will automatically retry.
  • Check Anthropic's status page at status.anthropic.com if it lasts longer than 10 minutes.
  • Do not restart your gateway. OpenClaw already handles retries for 529 errors with exponential backoff. Restarting just delays recovery.
Seeing 529 errors frequently? You can add a fallback model to your config so your bot switches to a lighter model when Opus is unavailable. Your ProductiveBot already includes a fallback configuration. Run openclaw config get agents.defaults.model to see your current fallback chain.

Still Not Working? Run These Commands

If you checked your billing, key, and tier and everything looks fine, try these steps on your Mac Mini. Open Terminal (press Cmd + Space, type "Terminal", press Enter).

Check your overall status

openclaw status

This gives you a full health check: gateway status, agent configuration, memory system, and connection health. Look for any items marked as errors or warnings.

Check the recent error logs

openclaw gateway logs

Look for the most recent error message. It will include the error code (401, 402, 429, or 529) and a short description of what went wrong.

Restart the gateway

openclaw gateway restart

If your bot was working and then stopped after an update or a power cycle, a restart often fixes it. This clears stale connections and picks up new config changes.

Update OpenClaw

npm update -g openclaw

Older versions of OpenClaw may not handle Opus 4.6 rate limit retries as gracefully. Updating ensures you have the latest retry logic, error handling, and model support.

Verify which model you are running

openclaw config get agents.defaults.model.primary

This should show anthropic/claude-opus-4-6. If it shows a different model, your configuration may have been changed.

How Much Does Opus 4.6 Actually Cost?

Claude Opus 4.6 is priced at $5 per million input tokens and $25 per million output tokens according to Anthropic's pricing page (as of March 2026). Here is what that translates to in real ProductiveBot usage.

Usage Pattern Messages/day Approx. Monthly Cost
Light (personal assistant, quick questions) 10 to 20 $10 to $30
Moderate (daily workflows, email drafts) 30 to 60 $30 to $75
Heavy (team use, automations, coding) 100+ $75 to $200+
Group chats (Discord, Telegram, Slack) Varies Can spike quickly

These numbers reflect Opus 4.6 pricing. A few things that increase cost:

  • Long conversations. Each message in a conversation includes the full chat history as context. A 50-message thread sends significantly more tokens than 50 separate one-off questions.
  • Tool-heavy responses. When your bot searches the web, reads files, or queries memory, each tool call is an additional API request with its own token cost.
  • Group chats. If your bot responds to every message in an active group, costs can spike fast. Consider adjusting your bot's group chat behavior if this is an issue.
Monitor your spending in real time. Go to console.anthropic.com/usage to see your daily and monthly API usage broken down by model. This is the best way to understand where your credits are going.

Skip API Billing Entirely

If you have a ChatGPT Plus, Pro, or Business subscription, you can connect it to your ProductiveBot through our Skill Store. Your bot uses your existing flat-rate subscription instead of per-message API credits. No balances to monitor, no expired credits, no rate limit tiers to think about.

Browse the Skill Store

Want to Switch Models?

ProductiveBot supports multiple AI models and providers. If Opus 4.6 is more than you need (or more than you want to pay for), you can switch to a lighter model or a different provider entirely. Here is how the options compare:

Option Billing Best for How to switch
Claude Opus 4.6 $5/$25 per MTok Complex tasks, coding, long context Current default
Claude Sonnet 4.6 $3/$15 per MTok Everyday tasks, 40% cheaper Change model in config
Claude Haiku 4.5 $1/$5 per MTok Quick tasks, 80% cheaper Change model in config
ChatGPT OAuth Flat monthly fee No API billing at all Install from Skill Store

To switch to a different Claude model, open Terminal and run:

openclaw config set agents.defaults.model.primary anthropic/claude-sonnet-4-6
openclaw gateway restart

Replace anthropic/claude-sonnet-4-6 with whatever model you want to use. Your bot will start using it immediately after the restart.

To switch to ChatGPT instead of Anthropic, visit the ProductiveBot Skill Store and install the Connect ChatGPT skill. It walks you through the entire process. Read more about why ChatGPT OAuth might be right for you.

Common Questions

My bot was working yesterday and stopped today. What happened?
The most likely cause is that your Anthropic credits ran out. Opus 4.6 costs more per message than smaller models, so balances deplete faster than you might expect. Log into console.anthropic.com and check your balance. If it is $0, add more credits and your bot will start responding again immediately.
I keep seeing "rate limit reached" but I barely use my bot
This is almost always a billing issue disguised as a rate limit error. When your credit balance is $0, Anthropic sometimes returns a rate limit error instead of a billing error. Check your balance first. If the balance is fine, check whether you have set a very low monthly spend limit on the Limits page. Also make sure you have not set a custom spend cap lower than what your bot actually needs.
I added credits but my bot still is not responding
Try restarting the gateway: openclaw gateway restart in Terminal. If it still does not work, check your API key at console.anthropic.com/settings/keys. Sometimes after a billing lapse, the API key needs to be re-verified or a new one needs to be created.
What is a "usage tier" and why does it matter?
Your usage tier determines how many requests your bot can make per minute and how much you can spend per month. Everyone starts at Tier 1 after purchasing $5 in credits. For Opus 4.6, Tier 1 allows 50 requests per minute and 30,000 input tokens per minute. You move up tiers automatically as you buy more credits. Most ProductiveBot users should aim for at least Tier 2 ($40 total purchased) for a comfortable experience.
Can I set a spending cap so I do not overspend?
Yes. On the Limits page, you can set a custom monthly spend limit. Just be aware that Opus 4.6 can use credits faster than lighter models. If you set the cap too low, your bot will stop working once it hits that cap. You will need to either raise it or wait until the next calendar month.
Do API credits expire?
Yes. Anthropic API credits expire one year after purchase. If you bought credits a long time ago and have not used them, they may have expired. Anthropic sends email notifications before credits expire, but it is easy to miss. Check the billing page for any expiry notices.
The error says "overloaded." Is my bot broken?
No. A 529 "overloaded" error means Anthropic's servers are temporarily busy. Opus 4.6 can be more affected by this than smaller models because it needs more compute. Your bot will automatically retry with exponential backoff. Do not restart your gateway. If it lasts more than 10 minutes, check status.anthropic.com.
Can I switch to a cheaper model?
Yes. You can switch to Claude Sonnet 4.6 ($3/$15 per MTok, about 40% cheaper) or Claude Haiku 4.5 ($1/$5 per MTok, about 80% cheaper). Run openclaw config set agents.defaults.model.primary anthropic/claude-sonnet-4-6 followed by openclaw gateway restart. Or you can skip API billing entirely by connecting your ChatGPT subscription through the Skill Store.
My bot sends one message and then stops. What is happening?
This usually means you hit a rate limit right after the first response. On Tier 1, Opus 4.6 is limited to 50 requests per minute and 30,000 input tokens per minute. A single bot response that uses tools (web search, memory, file reading) can generate 5 to 10 API calls and consume thousands of input tokens. Purchasing $40 total in credits automatically upgrades you to Tier 2 (1,000 RPM, 450,000 input tokens/min), which resolves this for almost everyone.
Can I use my bot with a different AI provider altogether?
Yes. ProductiveBot supports multiple providers. The Skill Store offers easy setup for connecting alternative models, including ChatGPT via OAuth (flat monthly fee, no API tokens). You can switch providers anytime without losing your data, memory, or settings. Talk to Scout if you need help choosing.

Quick Reference: Error Code Cheat Sheet

Code Meaning Your Fix
401 Invalid API key Create new key at console.anthropic.com, update with openclaw config set, restart gateway
402 Out of credits / payment failed Add credits at console.anthropic.com/settings/billing
429 Rate limit exceeded Wait 60 seconds. If frequent, buy $40+ total credits to reach Tier 2
529 Anthropic servers overloaded Wait 2-5 min. Not your problem. Do not restart gateway
500 Anthropic internal error Wait and retry. Check status.anthropic.com

Need help? Talk to Scout

Scout is our AI support assistant. It can walk you through each step, check your setup, and troubleshoot in real time.

Chat with Scout

You can also reply to your setup email or contact us at support@productivebot.ai

Reading next

Connect Your ChatGPT Subscription to ProductiveBot

Mar 18, 2026
Shopify API

How to Build a Skill for the ProductiveBot Skill Store

Mar 26, 2026
Shopify API

Leave a comment

This site is protected by hCaptcha and the hCaptcha Privacy Policy and Terms of Service apply.