Skip to main content
Troubleshooting and FAQs

'Concurrency Limit Reached' and '429 Too Many Requests'

What to do when calls are refused because every line is busy, or when the API or your provider answers 429 Too Many Requests.

Written By voiceaiwrapper

Last updated 26 days ago

These two messages look alike but have different causes. Concurrency limit reached means every call line on a Pod (an engine or provider connection) is busy, so a new call can't start. 429 Too Many Requests means something sent too many requests in a short time, either your system calling the VoiceAIWrapper API or VoiceAIWrapper calling your voice provider.

Use the first part of this article for call-line problems and the second for 429 errors.

Part 1: Concurrency limit reached

How call lines work

Each Pod has a Total number of lines: the most calls it can run at once. You split the total into Inbound lines (for incoming calls, web widget voice calls and browser test calls) and Outbound lines (for calls your campaigns and agents place). A call can only use a line from its own side, so a Pod can refuse an inbound call while outbound lines sit free. On VAW Native, web widget text chats don't use call lines; they have their own Website chats limit.

Where you'll see it

Where

What you see

VAW Native inbound phone call

The caller hears a busy signal. The call is logged with the end reason Concurrency limit reached.

VAW Native web widget voice call

The call doesn't start, and the widget tells the visitor it's unavailable. It's logged the same way.

VAW Native browser test call on an agent

"Concurrency limit reached β€” end an active call or test and try again."

VAW Native Call now on an agent's Deploy tab, or an outbound call started through the API

"All outbound lines are busy right now. Try again in a moment, or raise the Pod's outbound concurrency limit." Nothing is added to Call Logs.

Outbound campaign (any engine)

Nothing visible. The campaign waits and dials the next lead when an outbound line frees up. No leads are skipped.

VAW Native Call Logs hide refused calls by default. The call count says "missed calls are hidden". Filter by the end reason Concurrency limit reached to see them.

Fixes, most likely first

Your split gives too few lines to inbound. Move some outbound lines to inbound:

  1. In the Pod selector, click the pencil icon (Edit) on the Pod. The Update VoiceAI Pod drawer opens.

  2. Click Next to reach Set Concurrency. On Vapi, Retell AI and ElevenLabs Pods, this checks your keys with the provider again.

  3. Change Inbound or Outbound. The other field updates so the two add up to the Total, which you can't edit here.

  4. Click Update Concurrency.

Remember that web widget voice calls and browser test calls also use inbound lines. See Split Concurrency Between Inbound and Outbound Calls.

All outbound campaigns on a Pod share its outbound lines. Check:

  • The outbound share is big enough. Give outbound more lines in Set Concurrency if inbound has room to spare.

  • Campaign Priority (1-10). A campaign with a higher number gets more of the shared lines while several campaigns have leads waiting. Set it under the campaign's Settings β†’ Outbound behavior. If you don't see the field, your campaigns share lines without a priority setting.

  • Long calls hold lines. A campaign with long calls frees lines less often. Shorten the agent's maximum call length if calls run on.

See Outbound Dialing Rules: Attempts, Redial Wait, Voicemail and Priority.

On your own provider account, the provider has its own limit. If the Total in VoiceAIWrapper is higher than that limit, the provider refuses the extra calls even though VoiceAIWrapper shows free lines. If it's lower, you're leaving capacity unused.

Engine

How to bring the total in line

Vapi

Edit the Pod and set Vapi Concurrency Limit to the same number as your Vapi organization ("Please make sure this value is same as in your Vapi.ai org.").

Retell AI

On the VoiceAI Pods page, click Refresh next to Concurrency on the Pod's card ("Sync concurrency from provider"). If the total changed, the split resets to half and half, so check it afterwards.

ElevenLabs

The total is set from your ElevenLabs subscription tier when you connect. If you've changed tier since, contact support.

When Vapi refuses a campaign call because your Vapi organization is over its limit, the lead isn't marked invalid. It goes back in line and is tried again.

See Manage Pods: Rename, Update Keys, Switch and Delete.

Pay-as-you-go includes 6 concurrent calls and Growth includes 10. Additional concurrency on either plan costs $10/line/month or $100/line/year. Check your total in the Pod’s Set Concurrency step and contact support if you need help increasing it. For Scale, check your current plan allowance.

Check for calls still in progress: open test calls in other browser tabs, or calls that didn't hang up. The Live operations section of the Pod's dashboard (Pod Dashboard in the sidebar, or Dashboard if your only Pod is VAW Native) shows the live calls on the Pod right now. End anything still running, then try again. If lines stay busy with no live calls, contact support with the Pod name and the time.

Part 2: 429 Too Many Requests

A 429 status means a rate limit was hit. Work out who sent the request and who answered with 429.

Your system calling the VoiceAIWrapper API

The VoiceAIWrapper API limits how many requests each API key can send to each endpoint in a minute. Most endpoints allow 100 requests per minute. Over the limit, the API answers status 429 with:

{
  "message": "Rate limit exceeded. Please try again later.",
  "error": "RATE_LIMIT_EXCEEDED"
}

To stay under the limit:

  1. Use the bulk endpoints. One Add Bulk Leads request carries many leads but counts as one request. The same goes for bulk update and bulk remove. Sending one request per lead from a loop or an automation is the most common cause of 429s.

  2. Wait before retrying. After a 429, pause before you send again, and wait longer if it happens again. Retrying at once only adds to the count.

  3. Spread big jobs out. Send large imports in batches over several minutes.

The full list of endpoints and limits is in API Reference Overview and Rate Limits, and the bulk lead requests are shown in Leads API Quick Start. Rate-limited requests are refused before your API key is checked, so they aren't stored in API & Webhook Logs. Count your requests on your side, or check your automation tool's run history.

VoiceAIWrapper calling your voice provider

When a campaign places calls on your own Vapi, Retell AI or ElevenLabs account, the provider can answer 429 if your account is sending too many requests. You'll usually see this as a lead error that mentions 429 or "Too Many Requests".

What happens to the lead depends on your account:

  • The lead goes back in line and is tried again between 30 seconds and 2 minutes later. Its error says, for example, "Vapi rate limit (429). Will retry after 75s." Nothing else to do.

  • Or the lead is marked Invalid, with the provider's 429 error as the reason.

For invalid leads, fix the pressure on your provider account first, then put the leads back in line:

  1. Lower the outbound share in the Pod's Set Concurrency so fewer calls start at once, or make sure the Pod's total matches your provider limit (see Part 1).

  2. Check other tools using the same provider account. Another platform or automation using the same API key counts toward the same provider limit.

  3. Raise the limit with your provider if you regularly need more.

  4. Mark the leads valid: pause the campaign, select the leads on the Invalid tab, then Bulk Actions β†’ Mark Valid, and resume. See CSV Import Errors and Invalid Leads.

Related articles