Who can use Developer Agent?

Workday's own getting-started activities need only a free Developer Site account (created with a personal email address) and a development tenant. With those you chat with Developer Agent from the Ask Developer Agent box on the home page or build a custom agent with Build with Developer Agent.

The commercial side is separate. Workday's Developer Agent product page says it is available through Workday Extend Professional, will include Flex Credits-based pricing, and directs readers to their Workday account team for access and pricing. The developer documentation now calls that SKU Extend AI. The Rate Card gives that offer an annual allowance of 50,000 Developer Agent invocations; after that, each invocation costs 2 Flex Credits.

Where Developer Agent runs

Developer Agent works in three places, all against the same Workday account:

  • The Developer Site: the home page box, and the Chat with Developer Agent icon at the top right of any page.
  • App Builder: in an Extend app or an Agent app, where it proposes changes in Draft Mode that you preview in App Preview before you apply them.
  • Your coding tool: the Developer Agent Extension runs in Claude Code and VS Code-compatible editors. Workday lists Claude Code, Cursor, VS Code and Google Antigravity. It needs the Workday CLI (WDCLI) installed and signed in with wdcli auth login, and the two share one session. There you approve each proposed change in the editor, or turn on auto-approval, and pull an existing app with WDCLI before working on it.

For the coding-tool setup, see Developer Agent in Claude Code and Cursor and the WDCLI guide.

What your Developer Site role still decides

Developer Agent writes into apps, so your company role on the Developer Site still bounds what you do with the result. Developers and Company Administrators can promote apps to Production. Specialists can build and promote only to Implementation and Sandbox. Members have view-only access. Developer Agent also doesn't configure security: after you deploy an app or agent, security is set up in the tenant by someone with the right Workday domains.

Some generation needs a tenant connection. You must be connected to generate Graph queries for business objects (the objects must be in the app and deployed), to generate WQL (a non-Development tenant must be opted in to Innovation Services for Workday AI Gateway), and to generate agent skills, because that connection is how Developer Agent learns which agent-ready tools the tenant has.

What counts as one invocation

Workday defines a Developer Agent Invocation as a query or command submitted to Developer Agent in connection with development in Workday Extend, whether through the chat on the Developer Site or through a plugin or connector to a supported third-party coding tool. In practice:

  • By that definition, each prompt you send is an invocation, whether it is a question ("What's the difference between a synchronous and an asynchronous orchestration?") or an instruction ("Add a validation rule to SubmitExpense").
  • Read literally, a follow-up such as "tell me more" is another query and so another invocation. Workday does not spell this out, so check the count on the Entitlements dashboard after a test session.
  • Running the custom agent you built is not a Developer Agent invocation. That is a Custom Agent Invocation, metered on its own line by model tier, plus its tool calls. See why custom agents use their allowance quickly.

How the 50,000 allowance works

The Rate Card footnote for Developer Agent sets the rules:

  • Flex Credits are consumed only after the customer exceeds the annual entitlement of 50,000 invocations.
  • The entitlement resets on each anniversary of the subscription start date.
  • It is prorated to the subscription period's start and end dates, so a part-year term gets a part-year allowance.
  • Unused invocations expire at reset. They don't roll over.

After the allowance, the rate is 2 Flex Credits per invocation. The Rate Card states that its rates are informational as of October 12, 2026 and that the customer's agreement sets the final terms. Flex Credits themselves are a UMSA arrangement; the Flex Credits guide explains the agreement side. If price comes up with a client, the figure that binds them is in their order form, not the public card.

Where to see how many you have used

Workday names two places to monitor Developer Agent invocations:

  • The Entitlements dashboard on the Developer Site. Its Organization AI Usage entitlement shows the AI invocations used and the total available. Workday describes the dashboard as the place for administrators to view and manage Extend usage.
  • The Platform Consumption Console in the tenant, for UMSA subscribers. It shows Flex Credit usage by service, updated daily from 2:00 PM UTC for the previous day, and sends default alerts at 80%, 90% and 100% of the credit threshold to users with the Management Dashboard: Platform Consumption Console domain.

The Rate Card states the entitlement per customer, not per person, so developers working under one subscription draw on the same 50,000. A team building in parallel reaches it sooner than one consultant.

Making the allowance last

Developer Agent keeps the context of the conversation and remembers each iteration while you code, which is why a complete prompt costs less than a string of corrections. Workday's own guidance points the same way: treat a prompt as delegating to a colleague, with a clear objective, background and specific parameters. Two habits that follow from that:

  1. Write the requirement once, with field names, the business object and the expected behaviour, rather than correcting the output line by line.
  2. Ask questions about Workday APIs or concepts in the same thread where you are building, so the answer and the follow-up code come from one context.

Developer Agent's output still needs review. Its responses are not deterministic, and Workday's guidance is to review proposed changes before applying them and not to put personal data in prompts. A review checklist for Developer Agent output helps make that review quick.

Sources checked