Skip to content
KiweelyKiweelyKiweely

Workday API Rate Limits When Many Employees Ask Questions in Chat

Workday throttles REST and RaaS with HTTP 429 when a tenant is overloaded, and SOAP can surface the same pressure as HTTP 500. Extract ceilings and EIB concurrency are different meters. No page fetched here prints one requests-per-second number for employee chat.

  • Checked against Workday on
  • Applies to Workday Extend
  • 5 min read
  • Published
  • By Kiweely
On this page
Workday API Rate Limits When Many Employees Ask Questions in Chat

A company-wide assistant does not get its own Workday capacity. Every question becomes REST, WQL, SOAP, or a custom report on the same tenant that payroll and integrations already use. Workday publishes how it pushes back, and it publishes several ceilings that are easy to misread as a chat quota.

Kiweely Bridge (coming next) bills unlimited calls at the company price. That sentence is about Kiweely’s invoice. It does not raise the tenant limits below.

What happens when the tenant is busy

Integrations, EIB, and Web Service Limits says Workday throttles, under high load:

  • Workday Web Services (SOAP)
  • Reports-as-a-Service
  • The REST API

When other requests overload tenant resources, Workday rejects calls with:

API Response the page names
APIREST and RaaS Response the page names429 Too Many Requests
APISOAP and RaaS Response the page names500 Internal Service Error, which can mean too many requests

Workday’s advice on that page is an exponential back-off and caching. It also says a burst of invalid requests can be rate-limited with 429 so one customer’s error spike does not exhaust shared resources. The opening note says Workday may throttle or suspend access until usage is back inside the limit, and that extra capacity, where Workday sells it, is a separate purchase.

This page does not print a single “N requests per second” figure for ordinary REST. Do not copy a number from a vendor’s blog into a design review. The entitlements screen inside a tenant is the place a customer sees their own allowance; this public reference is the behavior when the tenant is already overloaded.

Ceilings that are not a chat quota

Same page, different meters. Quoting them without the label causes bad capacity plans.

API extract requests are RaaS extract interfaces only. The maximums are totals across those extract interfaces:

Customer size on the page Per 24 hours Per 60 minutes
Customer size on the pageFewer than 3,500 workers Per 24 hours75,000 Per 60 minutes15,000
Customer size on the pageFewer than 10,000 workers Per 24 hours100,000 Per 60 minutes30,000
Customer size on the pageFewer than 100,000 workers Per 24 hours250,000 Per 60 minutes75,000
Customer size on the pageMore than 100,000 workers Per 24 hours350,000 Per 60 minutes100,000

An employee asking “what is my time-off balance?” is not, by itself, an extract. An integration that pulls a report every few minutes is. If the assistant is wired to a web-service-enabled custom report and someone asks for the whole population, you are much closer to report and response limits than to a chatty lookup.

EIB concurrency is up to 5 concurrent integrations if resources are available. Workday suggests avoiding the top of the hour. That limit is for EIBs, not for how many employees may open a chat.

Other hard sizes that still apply if a tool returns a large payload:

Limit Figure on the page
LimitWeb-service response instances Figure on the page1 million. Narrow with criteria, filters, or dates
LimitRaaS advanced custom report enabled for web services Figure on the page2 GB
LimitIncoming web-service request, other than import Figure on the page500 MB
LimitIncoming import web-service request Figure on the page2 GB
LimitHTTP header size Figure on the page16 KB
LimitSOAP and REST attachment or image Figure on the page30 MB
LimitPaged-request cache Figure on the page2 hours between paged requests, and 30 minutes after the last page. Page 1 builds the cache and should not use a tight timeout

Workday does not log GET requests, unsuccessful requests, or web-service responses. A 429 storm can be invisible in the response log unless you send external integration HTTP headers for server-side tracking.

WQL limits that change how a question should be asked

WQL result limits say a query returns at most 1 million rows.

Where you call Timeout on that page
Where you callThe tenanted host Timeout on that page30 minutes
Where you callThe Workday Cloud Platform API gateway (api.workday.com) Timeout on that page5 minutes

If the query would exceed 1 million rows and you set no LIMIT, it fails. A LIMIT of 1 million or less succeeds and returns at most 1 million rows. You view 10,000 rows at a time. LIMIT as a clause caps the total. limit as a parameter sets the page size. The keyword is not case-sensitive; Workday uses both forms.

The WQL REST reference adds the transport split: GET /data for a query under 2,048 characters, POST /data for a query from 2,048 to 16,000 characters. You only receive rows your security can see. Tenanted hosts use https://{hostname}/api/wql/{version}/{tenant}. The platform gateway uses https://api.workday.com/wql/{version}.

A headcount question that scans every worker, with no organization filter, is how a chat hits the row cap and the timeout. A question about the signed-in worker does not.

How to plan the busy morning

Open enrollment, payday, and a manager cascade are the real tests. Workday’s published controls are tenant-wide. Practical consequences, without inventing a rate:

  1. Keep the default question small. The person’s own balances and elections are a different load from an unconstrained report.
  2. Cache reference data that does not change per employee, as Workday recommends, and back off on 429 and on SOAP 500 that is really load.
  3. Do not schedule a full-population extract at the same minute everyone is paid.
  4. Treat page 1 of a paged call as slow. Give it time to fill the cache.
  5. Separate “Kiweely does not meter calls” from “Workday will not throttle.” Bridge (coming next) is the first sentence. The limits page is the second.

A personal developer tenant is free for one person on Bridge and is the wrong place to learn a 5,000-person spike. Load still belongs on a non-production tenant that has production-like volume, and even there the limits page allows Workday to throttle.

Frequently asked questions

Will 1,000 employees chatting at once get 429?

Workday does not publish the concurrency number that would answer yes or no. It does say REST and RaaS return 429 when other requests overload the tenant, and SOAP can return 500 for the same reason. Plan for that response. Do not plan for an unpublished quota.

Does unlimited Bridge pricing remove these limits?

No. Unlimited calls is the company invoice for Kiweely Bridge, which is coming next. The tenant limits in this guide still belong to Workday.

Sources checked

Related guides

All guides

Let Workday check the work.

Kiweely writes WQL and builds orchestrations in your own tenant, and Workday, not Kiweely, says whether they passed.

Get the next guide in your inbox.

One email a month with new guides and one Extend tip. Nothing else.

We’ll email you once to confirm. Unsubscribe in one click from any issue.