How to Connect Workday to Claude with the MCP Connector

Introduction

On October 12, 2026, Workday and Anthropic announced a strategic partnership at Workday Rising 2026 that connects Claude’s reasoning to Workday’s HR and finance data through the Model Context Protocol. The product is called Sana in Claude: employees can find Workday information, take action, and approve work without leaving Claude, while every request still follows the roles, permissions, approval chains, and audit controls already configured in Workday. This tutorial is for Workday admins evaluating the integration and for employees at enrolled organizations who want to understand how the connection is set up, what it can do on day one, and where its guardrails are. Because the connector is in early access, this guide sticks to what Workday and Anthropic have officially confirmed and marks the standard Claude connector steps clearly.

Quick answer

Workday connects to Claude via Sana in Claude, an MCP connector announced October 12, 2026. A Workday admin enables early access once per organization; employees sign in with Workday SSO, and every action follows existing Workday permissions and approval chains.

Evidence

The scale of this integration is what makes it worth tracking. Workday states that more than 11,500 organizations worldwide, including more than 65% of the Fortune 500, trust its platform for HR, finance, and IT, as of the October 2026 announcement. That installed base is the reason a governed MCP connector matters: it brings agentic AI to systems where the margin for error is effectively zero.

The technical design is explicit about governance. Claude and Sana exchange information and tasks through the Model Context Protocol, an open standard, so work carries forward rather than starting over when you move between the two. Workday’s announcement headline states the rule plainly: Workday’s permissions and business rules apply to every request, and the user confirms every change. For admins, the most practical line in the release is that admins connect once, with no new permission model to build and no per-user setup.

Gerrit Kazmaier, Workday’s president of product and technology, framed the intent this way: “Customers want the best domain AI agents for both their people and money that understand the business rules they run on and leverage the extraordinary reasoning in Claude.” The trust signal behind the partnership is concrete, too: Anthropic itself runs its finance and people operations on Workday Financials and Workday HCM, under the same governance, rules, approvals, and audit controls it sells to customers. These details come from the official Workday and Anthropic partnership announcement, with the product rollout context in Workday’s Rising 2026 product announcements. The protocol underneath is documented at the Model Context Protocol open standard site. If you are new to the protocol itself, our primer on how the Model Context Protocol exposes tools explains the client-server model these connectors rely on.

Prerequisites

• A Workday tenant with an admin who can enroll the organization in early access programs, plus a Workday account executive or support contact.
• A Claude Team or Enterprise plan, with access to a Claude org owner or primary owner (connectors are managed at the organization level).
• Workday single sign-on (SSO) working for the employees who will use the connector.
• Familiarity with your Workday security groups and approval chains, since those exact rules will govern what Claude can do.
• About 30 minutes for the admin side, plus 10 minutes per employee for first sign-in.

Step-by-step guide

Step 1: Confirm your organization is eligible

Sana in Claude is available to early access customers as of October 2026 and is expected to reach general availability in late 2026. Eligibility therefore has two gates: you must be a joint Workday and Claude customer, and your Workday tenant must be enrolled in the early access program. Ask your Workday account team to confirm enrollment before you touch any settings. If your tenant is not enrolled, nothing on the Claude side will show the connector, and that is the expected behavior, not a bug.

Step 2: The Workday admin enables the integration once

The design goal Workday states is “admins connect once.” In practice this means the enablement happens at the tenant level in Workday, not per user. Your Workday admin enrolls the tenant in the Sana in Claude early access program through the usual early-access channel (account team or admin console program enrollment). Once enabled, the existing Workday security model carries over automatically: the roles, domain security policies, approval chains, and audit logging you already maintain become the rules Claude operates under. There is no parallel permission model to build, which is the single biggest admin time-saver in this integration.

Step 3: The Claude org owner adds the connector

On the Claude side, connectors are an organization-level feature. A Claude owner or primary owner opens Settings, then Connectors, and adds the Workday connector for the whole organization. This mirrors the standard documented flow for Claude connectors: the admin adds it once, and it becomes available to members, who then authenticate individually. Individual employees cannot self-serve this step on a Team or Enterprise plan; if you do not see the connector, your org owner has not added it yet.

Step 4: Employees sign in with Workday SSO

After the connector is available, each employee opens it in Claude and signs in with their Workday credentials through SSO. This per-user sign-in is what binds Claude’s requests to that specific person’s Workday identity, and therefore to their permissions. If your organization uses an identity provider such as Entra ID or Okta in front of Workday, the sign-in follows that same flow, including any conditional access policies. A successful sign-in does not grant new powers; it only lets Claude act with the powers the employee already has.

Step 5: Run your first governed workflows

Workday’s announcement documents two concrete day-one scenarios, and they make good first tests because each exercises a different governance path:

1. HR scenario. A hiring manager asks Claude to compare open roles against existing headcount and budget, then build an internal mobility plan that fills those roles with people already on the team. This is read-heavy: Claude analyzes Workday HCM data and drafts a plan, surfacing only what that manager is permitted to see.
2. Finance scenario. A finance leader asks Claude to check invoice exceptions against purchase orders, then approve them in Claude or route them to Sana. This exercises the write path: the approval still follows the customer’s existing approval chain, and the user confirms every change.

Start with read-only questions (“show my open expense reports”, “summarize headcount changes on my team this quarter”), verify the answers match what you see in Workday, then try one approval-gated action and confirm the approval request appears in your normal Workday inbox.

Step 6: Verify governance is actually working

Do a deliberate negative test: ask Claude for something you know you are not permitted to see, such as another team’s compensation data. The correct behavior is a refusal or a visibly limited answer, because Claude surfaces only what each person is permitted to see. Then check your Workday audit trail for the actions you took through Claude; they should appear like any other Workday activity, attributable to your user. If either check fails, disconnect the connector and escalate to your Workday admin before continuing.

Troubleshooting

The connector does not appear in Claude’s connector list. The most common cause is that your organization is not enrolled in early access, or the Claude org owner has not added the connector yet. Confirm enrollment with your Workday account team first, then confirm with your Claude org owner. On plans without org-level connectors, the option will not appear at all.

Sign-in loops or “access denied” during Workday SSO. This points at identity, not at Claude. Verify the employee exists and is active in the Workday tenant, that SSO is correctly configured, and that any conditional access policy in your identity provider permits the sign-in. The Claude connector cannot authenticate a user that Workday itself does not recognize.

Claude refuses an action or routes it for approval instead of completing it. This is the governance working as designed, not a malfunction. Workday’s permissions and business rules apply to every request, and approval chains still fire. Complete the approval through your normal Workday or Sana inbox. If you believe the refusal is wrong, the fix is in Workday’s security groups, not in Claude.

The connector shows as connected but returns no data. The org-level connection is likely incomplete: the Workday admin’s one-time enablement may not have finished propagating, or the tenant’s early access enrollment lapsed. Have the admin re-verify enrollment and the connector status in the admin console. Also confirm the employee completed their individual SSO sign-in; an added-but-never-signed-in connector shows as available but cannot query anything.

FAQ

Is the Workday to Claude MCP connector generally available?

Not yet. Sana in Claude is available to early access customers as of October 2026, with general availability expected in late 2026, according to Workday’s official announcement. Your Workday admin or account team can confirm whether your organization is enrolled.

Does Claude get a copy of our Workday data when we connect it?

No. The integration is designed so Workday permissions, approval chains, and audit controls apply to every request, and the user confirms every change. When Sana calls on Claude’s reasoning for a step, it shares only the context that step needs, and exceptions go to a person for review.

Do we need to rebuild our permission model for Claude?

No. Workday states that admins connect once, with no new permission model to build and no per-user setup. Claude surfaces only what each person is already permitted to see in Workday.

Which Claude plan do we need for the connector?

Sana in Claude is an enterprise integration for joint Workday and Claude customers. On the Claude side, connectors are managed at the organization level on Team and Enterprise plans, so a Claude org owner or primary owner must enable it before individual employees can use it.

Can I use the Workday connector from Claude Desktop or Claude Code?

The announcement describes the experience living in Claude and in Sana. For other MCP clients, the general pattern is the same as any remote MCP server: add the server URL as a connector and authenticate with OAuth. Our walkthrough of connecting MCP servers to Claude Desktop covers that generic flow step by step.

Conclusion

Connecting Workday to Claude via Sana in Claude is an admin-led, one-time enablement followed by per-employee SSO sign-in, with your existing Workday permissions, approval chains, and audit trail governing everything Claude does. Start with the two vendor-documented scenarios, headcount and budget analysis on the HR side and invoice exception approvals on the finance side, and verify the guardrails with a deliberate negative test before rolling out widely. If you want to understand the protocol underneath first, read our Model Context Protocol primer next. For a self-hosted MCP server that queries business data instead of HR data, see our Business Central MCP server tutorial.

Frequently asked questions