Bug: Facing FlowDisabled error during test run in Copilot Studio Kit

[aishwaryamohanty2105]

Description

When user is executing any test case in Copilot Studio Kit, most of the test cases are failing with ErrorCode: FlowDisabled. The same test cases work fine when user enters them manually in Copilot Studio but fails in kit.
The issue if faced by multiple users and is impacting our deliverables.

Steps to reproduce

1. Configure copilot agent to the kit.
2. Create test set
3. Upload test cases
4. Run the test cases

Component_Test (L2).csv

Expected behavior

Expected responses: User should get proper response similar to manual tests

Actual behavior

Actual responses: User is getting FlowDisabled error for most of the test cases

Environment

App name: Copilot Studio Kit
Organization name: ACN_COPILOT_STAGE_PER-1608_PeopleAssist_315447
Channel: Monthly - June 2026 (app)
SANDBOX
Server version: 9.2.26054.00154
Client version: 1.4.12088-2606.1
Session ID: 0757e607-3b8a-44a3-acc2-51b68e8cac13
Time: 7/7/2026 7:11:01 PM GMT+0530 (India Standard Time)

Additional context or logs

[
{
"text": "An error has occurred.\nError code: FlowDisabled\nConversation Id: d4a1f855-87fa-4a85-9c04-b168a6125d5c\nTime (UTC): 2026-07-07T04:39:57.565Z."
}
]

[SiddharthPoweCat]

Hi aishwaryamohanty2105 — thanks for the report, and apologies for the delay.

What this error actually is

FlowDisabled is not generated by the Copilot Agent Kit. It is a Copilot Studio runtime error returned by your agent itself, and the Kit simply records the agent's reply verbatim in the test result. Copilot Studio emits it when a topic reaches a Call an action step that invokes a Power Automate / agent flow which is not in a running (On) state.

Why it works manually but fails in the Kit

This is the key difference, and it's expected behavior rather than a defect:

	Copilot Studio Test pane	Copilot Agent Kit
Version executed	Draft / authoring copy	Published agent
Transport	In-product test harness	Direct Line (token endpoint or channel secret)
Identity	You, the maker	Unauthenticated or configured Kit identity

So "passes manually, fails in the Kit" almost always means the published agent — or the environment it is published into — has flows that are turned off, or the Kit is pointed at a different agent/environment than the one you're testing by hand.

Quickest isolation test

Open your agent in Copilot Studio → Channels → Demo website, and run one of the failing utterances there. The Demo website uses the same published + Direct Line path as the Kit.

• Fails there too → confirmed: this is an agent/flow configuration issue, not a Kit issue. Continue below.
• Works there but fails in the Kit → the Agent Configuration is pointing somewhere else. Re-copy the Token Endpoint (Settings → Channels → Mobile app / Email) or the Direct Line secret into the Agent Configuration record and re-run.

Steps to resolve

1. Look up the failed conversation. You have Conversation Id: d4a1f855-87fa-4a85-9c04-b168a6125d5c. In Copilot Studio go to Analytics → Sessions, find that session, and open the transcript. It will name the exact topic and action that failed — that identifies the offending flow directly.

2. Turn the flow(s) back on. In make.powerautomate.com, in the same environment as the agent, open Solutions → (your solution) → Cloud flows and confirm each flow called by the failing topics shows status On.

   This is very likely your root cause. Your environment is listed as ACN_COPILOT_STAGE_PER-1608_PeopleAssist_315447 (SANDBOX). Sandbox environments that have been copied, restored, or reset come back with all cloud flows turned off, which would explain why most of your test cases fail at once while the draft/Test-pane path still behaves.

3. Check for auto-suspension. Power Automate also disables flows when:

   • the flow owner's Power Automate / Power Apps license was removed or expired,
   • the flow accumulated repeated failures (you'd have received a "your flow has been turned off" email),
   • a connection is broken — look for Fix connection on the flow's detail page.

4. Republish the agent after enabling the flows, then re-run a single test case to confirm before re-running the full set.

5. Verify DLP. Confirm the connectors used by your flows and by the Kit are allowed in the environment's DLP policy — see TROUBLESHOOT.md → AppForbidden or other DLP errors.

Could you share the following?

So we can confirm and rule out anything Kit-side:

1. Copilot Agent Kit solution version — Power Apps → Solutions → Copilot Agent Kit (or Copilot Studio Accelerator on older installs) → the Version value, e.g. 20260708.6.
2. Result of the Demo website test in the isolation step above.
3. Whether the flows called by the failing topics are On in the agent's environment.
4. The session transcript screenshot for the conversation ID above.

If you'd like to walk through it live, you're welcome to join Office Hours