Agent Studio

Hi everyone,I’m building an Agent Studio plugin that uses a Compound Action to log and update employee timesheets via an external API.Our middleware API is designed to return detailed business and client validation errors (e.g., missing comments, invalid locations, closed periods) using standard HTTP status codes like 400, 422 and 500. Crucially, the exact reason for the failure is provided inside the JSON response payload, like this:{  "status": 422,  "error": "Business Rule Failed",  "message": "The following charge requires a justification comment."}The Problem:When the HTTP Action receives a 4xx or 5xx status code, Moveworks correctly treats it as an error and routes execution to my catch block. However, the platform appears to drop the actual JSON payload body during the crash. Because the payload is not exposed to the Data Bank in the catch block, I am unable to extract the dynamic "message" field to display to the user.Is there any way in Agent Studio to access the raw JSON resp

Has anybody who has set up the Jira (Live) Setup for Enterprise Search successfully gotten to a point where end-users can validate using OAuth? The documentation doesn’t demonstrate how a typical end-user might authenticate to then utilize the full Jira Live-Search functionality as described. 

Hi Team,When the payload contains a URL that uses a protocol other than HTTPS or HTTP in the Notify, the URL is broken in the frontend user message. For example, when we send a URL with a different protocol (such as mailto:, msteams:, or a custom protocol), the link breaks and is not rendered correctly. Sample payload with a non-HTTPS protocol:Bot response: HTTPS protocol with the same structure: Bot response: Please help us find a solution for this issue, as it is currently impacting multiple live use cases.​@abhiagarwl  - FYA

I am currently working on a use case to look up assigned training for a coworker. However, I am running into an issue with the query provided in the Moveworks documentation.The problem is that an assigned training course does not appear in the HTTP action results until the user manually enrolls in the course by clicking it in the Workday portal. Even though the training is already assigned and visible to the user as something they need to complete, it is not returned by the query until after they interact with it.Is there another query or endpoint that I can use to retrieve all assigned training courses for a user without requiring them to first click or enroll in the training within Workday? My goal is to return all assigned training, including courses that have not yet been opened or interacted with by the user.Use Case: Plugin | Look up Assigned TrainingsAPI: /ccx/api/wql/v1/{{tenat}}/dataBody: {  "query": "SELECT enrolledContent as Course, learner, registrationStatus, completionSta

In my application, all request bodies need to be form data with data as the key and the json body as the value. I am able to handle this in http actions with raw json body by having it be something like data={<insert body here>} and having the header of Content-Type=application/x-www-form-urlencodedI want to use the new data mapper functionality in http actions so it will handle escaping and such for me, however I cannot seem to find a way to get it to work with this setup. I have attempted numerous times with variations of data: <json to mapper syntax> , but the app always returns with a missing data element error. Is there any way to handle this currently?

Hi everyone,I am configuring deduplication for a Moveworks webhook Listener and need to deduplicate by `event_id` from the parsed payload. The listener documentation shows that incoming webhook bodies are available as `parsed_body`.My payload looks like this:```json{   "Report_Entry": [      { "event_id": "ABDCS",         "...": "..."      }   ]}

Hi,When deploying a use case from one instance to another, the system consistently prompts for the recreation of shared action / compound actions, displaying the message: “An object with the same name already exists.” This necessitates either recreating the object or reconfiguring the pre-existing object and performing cleanup activities for newly created action/compound action at a later stage after the use case is configured.This approach increases the risk of human error and may lead to unintentional changes being made directly in the production environment, which is not aligned with established process compliance standards.it is a know issue / what is the best way of deploying use cases on prod Regards,Rajeev 

Hi all,I'm running into a persistent "There was an error saving your change" dialog with Error Code: 4 when trying to save the new credentials in the Workday Built-in connector in Moveworks Sandbox Setup. Hoping someone has seen this before or can point me to the right fix.  

Installing a plugin used to be messier than it needed to be.If the target org already had a connector, HTTP action, compound action, or conversational action with the same name, Agent Studio created forced you to create another copy by renaming it. One install could leave you with Create Ticket, Create Ticket (1), Create Ticket (2), then a bunch of manual cleanup to figure out which one was the real thing.That changes today. Reuse Existing Assets on Plugin Install is now GA for all customers.When you install a plugin from Marketplace or import one from another environment, Agent Studio now checks for matching assets by name in the target org. For each match, you can choose:Incoming asset ChoiceServiceNow connector Use existing or configure newCreate ticket action Use existing or configure newNotify user action Use existing or configure newPick Use existing and the plugin points at the asset already in that org. Anything new still gets created as new.What this fixes:→ Le

We are using Moveworks to perform an AIOPs POC (& later implementation). I have created a new conversational process (validated and published) but when I am trying to configure a plugin (conversational one) I am not able to see this newly created (even I did log-out and refreshed page and even tried after a few hours). So, I am not able to attach this newly created conversational process to a conversational body within Plugin.

I am getting an error when trying to test any compound action in agent studio using the test functionality. It does not process anything, just returns:Message: Failed to process custom actionError Code: 4I have confirmed this in both production and sandbox for multiple compound actions that are live.

Hi,I am  integrating with Workday using an OAuth2 connector and are hitting a recurring lockout I can't pin down. Hoping someone has seen this pattern.Setup:- Auth Config: OAuth2- Grant Type: Refresh Token Grant- Credentials in use: client ID, client secret, refresh tokenWhat happened:1. We were getting "permission denied" on one new specific REST endpoint (GET /workers/{ID}/directReports) while all other Workday API calls worked fine.2. Working with our Workday team, we rotated the connector credentials — new client ID, new client secret, new refresh token — and the Workday team also made some configuration changes on their side.3. After that, everything worked, including the directReports endpoint.4. But a few hours later all the APIs failed. The API client was locked in Workday and we got:   {     "error": "invalid_client"   }5. The Workday team unlocked it, everything worked again, and then it locked again after about an hour. This has now repeated several times.Questions:- Has any

Looking to enable external connectors to Microsoft support (Copilot settings) and Adobe (Photoshop) related. Where do I enable these ones? Like I’m looking to retrieve data from external sources not internal ones. 

I am attempting to create a plugin in Agent Studio, followed the steps to first create a conversational process and clicked “Publish”. I confirmed that the Conversational Process I just created shows “Status: Published”. When I’m creating a Plugin and trying to select the Conversational process, however, I am unable to find it in the list. I searched by the first few characters of the Conversational Process name “rypasca” as shown in the screenshot below, and it still says “no options available”.Does anyone know what might cause this? I also saw this similar post here from a user: Newly Created Conversational process is not shown in the options in the plugin while trying to map newly created process to a conversational body | Community

I am attempting to test an HTTP action but receive the following error message: Status Code: -1Response:Invalid URL '': No scheme supplied. Perhaps you meant https://? I have the base url from my connector as https://us261.dayforcehcm.com and the endpoint as /Api/fcf/V1/EmployeeBalancePeriods Please let me know if anyone has an idea of why this might be occuring, or if additional information is needed from my end.  

When installing plugins from the Moveworks Marketplace, we're encountering Error Code 13121 which prevents successful installation. The error indicates a missing value in Workflow Configurations > Workflow Arguments.Problem SummaryIssue Type: Plugin Installation FailureSource: Moveworks MarketplaceAffected Area: Agent Studio / Plugin WorkspaceError Code: 13121Impact: Plugin cannot be installed, workflows remain incompleteError DetailsError Message: There are errors with the field: Workflow Configurations > Workflow Arguments [arguments[0].type_id] - value must be setError Code: 13121Trace ID: qRkjwu_cbqsTWU-U 

I have created a custom plugin in Moveworks for the ServiceNow Problem table. In this plugin, I am fetching details such as the Problem Record Number, State, and Priority from ServiceNow.However, I am facing an issue with choice fields like State and Priority. Instead of receiving the display label (for example, “Closed”), Moveworks is returning the backend value (for example, 7 for Closed).Example:In ServiceNow: State = Closed Backend value = 7 In Moveworks response: State = 7 (instead of “Closed”) Expected Behavior:I want Moveworks Assistant to display the user-friendly label (e.g., “Closed”) instead of the numeric backend value for choice fields.

Has anyone built a conversational workflow in Moveworks that leverages more than 1 file as an input (slot)? How are you achieving this as there is currently a limitation on 1 file per slot as mentioned in this blog post: https://www.moveworks.com/us/en/resources/blog/file-slots-file-uploads#one-file-per-slot-no-overwriting  Right now I am building additional slots for each file I need as an input, but it is making my conversational workflow very long. Being able to input an array of files would be great (ie. bulk uploading expense report receipts)

Hi ,I am working on PTO Workday Plugins.Today, I noticed an issue in Moveworks  like a hallucination / stale response behavior. Scenario:- Today, I asked Moveworks Bot: ""Give me time off requests". June 29th time off was in Approved status.- I cancelled the approved time off via MW Bot→ status became Approved (Cancellation Pending).- Then I cancelled the cancellation request directly in Workday → status should revert to Approved.- However, when I asked MW Bot: "Give me time off requests", it still showed Approved (Cancellation Pending). It seems MW Bot may be generating the response from previous conversation context instead of invoking the latest plugin (time-off list retrieval). Questions:- How can we enforce plugin execution (fresh data fetch) for queries like this?- Are there best practices to avoid stale or cached responses in such scenarios?

Hello All, I have created a custom plugin in Moveworks Sandbox environment and also published that pluginBut I’m unable to see in the Marketplace to run the plugin Please suggest me how to run the plugin and develop a bot like solution?

I am trying to send a notify message with a mailto link as part of the template but the link isn’t rendering properly. Trying to understand why. Other links in the message are rendering properly. I thought it may be the ** in the hyperlink text but its the same result without it. I saw another post mention mailto so I thought this should work but their approach was with the message coming from a variable in HTML and being manipulated back to the normal text whereas I want the template to include it each time as normal. Agent Architect also says that the normal markup hyperlink format should work.   This is the output.  

I wanted to share a useful architecture for empowering non-technical teams—like Sales, Finance, or Operations—to query complex databases using natural language. The logic is system-agnostic and can be applied to any SQL-based data source.The Problem: The "SQL Gap"In many organizations, critical business data is locked behind complex SQL databases. Non-technical users often have to wait for analysts, navigate clunky BI tools, or attempt to write SQL themselves (leading to errors or inefficiency).The Solution: Natural Language to SQLThis architecture allows users to ask questions like "What was the total revenue for Customer X in Q3?" or "Show me the top 5 products by volume last year"  and receive real-time data directly in MW Assistant. The plugin automates the translation of intent into an executable SQL query.Architecture Overview1. Data Modeling & TrainingThe core of the solution is a high-quality system prompt. Rather than just listing table names, provide a Data Model that inc

I am currently working in Agent Studio to create custom plugins and have implemented user input using slots. However, I am encountering an issue where all slot-related questions are presented to the user simultaneously.My requirement is to have the questions asked in a sequential manner — that is, the system should prompt the first question, wait for the user’s response, and then proceed to ask the next question, and so on.Could you please advise how to configure or design the slots so that they are prompted one at a time instead of all at once?for the reference I have attached the screenshot as well.