Agent Studio

 We are unable to leverage the output mapping in a way that provides consistent results to our users. Our use case is extremely simple. We are connected to a ServiceNow app for tracking issues. The sysparm_query pulls in a record number, il_status, severity, description (issue_detail), opened_by, assignee (responsible_user), and resolution (going_to_do). The testing and validation of the http action passed as expected. Using simple mode for the action activity within our conversational process returns inconsistent results so we switched to data mapper. We are unable to return any reasonably formatted response. Below is the Response Schema from my HTTP Action and then the Data Mapper. Any thoughts on what we are doing wrong? This is preventing us from moving forward with our plugin.Response Schema:{ "type": "object", "properties": { "result": { "type": "array", "items": { "typ

I am trying to build assign ServiceNow delegate use case, which is available in moveworks gallery, in this case i am trying to use user consent flow for all the connection that is authorization code user flow. First time attempt, I was expecting bot to serve a link to connection link to allow it but instead it loggin an error and not serving link to consent. I feel the connection which are the part of slot resolvers might not serve consent flow, i do have other usecases, and its working expected. but not in this case.any idea? you can reproduce the issue by using the plugin. 

I am trying to configure a webhook that asks the user for catering or room setup details (if required) when they book a meeting in Outlook.I have the following compound action, to be triggered from my Webhook, but it is telling me (as it also does in the documentation), that I need to publish my conversation as a plugin.steps: - action: action_name: mw.get_user_by_email output_key: catering_user input_args: user_email: data.meeting_room_catering.user_email - notify: output_key: catering_output recipient_id: data.catering_user.user.id message: RENDER(): args: meeting_room: data.meeting_room_catering.details template: "You have recently booked meeting room {{meeting_room}}. Do you require catering or help setting up the meeting room" actions: - key: nothanks

hi all, i’ve built a plugin based around clearing all of a user’s MFA methods attached to their IdP account. because of the some of the endpoints i’m querying, the responses take too long and i was getting a timeout error every time i prompted for this plugin at a certain step. i found this post that helped point me in the direction of running the actions that were timing out as compound actions instead (yay community gainz), then i set a return for the output. the plugin now functions all the way through which is great! however, i’ve noticed that when a compound action runs, the bot automatically sends this message “the process is running, i’ll let you know when it’s finished” except much more verbose, i’m assuming because the compound action runs async.my problem is i have 3 compound action

How can I update a use case in the new Agent Studio to handle fuzzy searching?Right now, it works fine if everything is spelled correctly, but if someone makes a typo, it doesn’t work.For example, I my bot, asked "Who is the PLM for XMD NAFTA" and got this back from the bot “I understand you are looking for the PLM contact for XMD NAFTA. I searched the PLM contact directory and relevant knowledge sources, but I was unable to retrieve the contact information for the XMD NAFTA PLM. Based on info you provided, I am unsure of how to help. Please click on Get Help below to review the supported help options.”However, there is no XMD NAFTA in the SharePoint list—it only has options like XD NAFTA, XMD EMEA, XM NAFTA, CLS NAFTA, FOX NAFTA, FOC NAFTA, DC NAFTA, PNP NAFTA, CAB NAFTA, FOD NAFTA, FOE NAFTA, and an LT NAFTA.I was expecting the bot to tell me "XMD NAFTA isn’t found" but also show a list of the closest matches so I can pick one.

Hi everyone,Apologies for the ignorance here—I honestly feel like I should know this already, but I get a bit stuck with output mapping. I’m the analyst working on our company's bot, and while we have an engineer on the team, I help out with some of the use cases. This is more of his role and area of expertise.Is there a straightforward way to create output mappers? I've tried using the DSL and Data Mapper Playground, and I've read through the Data Mapper / JSON Bender Reference and Moveworks Data Mapper Common Examples, but I haven't found them very helpful.Am I missing a simple trick here? I tend to prefer step-by-step instructions, so the guides that just provide general info are a bit hard for me to follow. If they don’t give us step 1—do this, step 2—do this thing, step 3—you need this it makes it hard for me.Thanks in advance for any tips or straightforward steps you can share to help me improve my efficiency with this. If you could break it down in

Hi Community,As discussed in the “AI Agent Marketplace Update - September 15th, 2025”.I have installed the plugin and for the most part it works great, accept in some cases the “Caller” value (User) is printing in the bot as hyperlink to the Snow sys_user table.If I then tell the bot to exclude the hyperlinked value for “Caller” straight away after the first utterance it the prints the “Caller” value as I would expect.Has anyone else installed this plugin and experienced the issue? If so have you been able to resolve the bug?@Moveworks Team - I have log a developer support ticket.Thank you.

Hello, We seem to be having a weird bug with confirmation policy and context clearing. We have a plugin that has a confirmation policy and in testing we did the following: We ignored the confirmation ask of the plugin and cleared context and when we pressed Yes to clear the context it cleared the context but also processed the Yes of the plugin and went through the plugin’s process We also tried canceling the request to the plugin then cleared context and when we pressed Yes on clearing the policy it also processed the Yes on the plugin and went through the process. Not sure if there is an issue with our context clear functionality because on another instance I cleared context and asked a question about something in the previous context and it still was able to recall the information. This plugin that we are building is for resetting an MFA and we can’t have the plugin being called to remove their M

Hi everyone,We’re exploring how to implement MFA reset functionality using Microsoft Graph APIs. The process involves two main API endpoints:List Authentication Methods API: To retrieve all MFA methods registered for a user. This gives visibility into their current devices. Delete Microsoft Authenticator Method API: To reset specific MFA methods. This is important because users may have multiple devices (like Device A, B, C, D). We want to allow resetting individual devices rather than removing all MFA factors. Users can choose which devices to reset or reset all if needed.Permissions and User Identity Verification:We plan to use delegated permissions with User Consent Authentication (OAuth 2.0 Authorization Code flow). This ensures the user explicitly consents, adding security. We will also

Hello,I have a cmdb lookup skill which returns the business application list, I am been trying to instruct the LLM to display only name, description and application owner. tried below map code also tested on data mapper playground, but getting response saying no records. without below code, it does display all the results. when there is single result, i would like to show something like below, but no much information on logs and no failures also bot is saying just ,,”No ownership information was found in the business application database or internal knowledge sources.”please help me to identify the issue. also give me a good documentation back end supported syntaxes like system_prompt, display_instructions, I want to study deep here.  {  "result": [    {      "short_description"

The recent Agent Studio Updates & Next Steps email mentions Agent Studio (Classic) will be retiring in the H1 2026. Will Events be a part of this retirement or will they move to Agent Studio? If so, are Webhooks the intended replacement for Events?We use Events in many messaging campaigns and would like to plan for the future. 

Is it possible to customize ticket resolved ticket notification to include a link to a survey? Or what is the recommended way to build this out using Agent Studio to serve up the survey?

My team is working on a Databricks SQL statement and wanted to see if anyone may have any insights. It looks like our bot isn't showing the responses when we test. The logs show that the actions are running without errors, but the conversation logs show that the output is coming through as null. Since the mapping is Moveworks-specific and a bit complex, I’ve attached the logs so others can take a closer look and help troubleshoot. Statement is something like this, I’ve redacted proprietary information:{  "statement": "SELECT DISTINCT <InvoiceID>, <SupplierName>, CONCAT(<BusinessUnit_PO>, ' ', <PurchaseOrder>) AS PO_ID, <VoucherStatus>, CASE WHEN <VoucherStatus> = 'STATUS1' THEN <PaymentDate> ELSE NULL END AS PaymentDate, CASE WHEN <VoucherStatus> = 'STATUS1' THEN <PaymentID> ELSE NULL END AS PaymentID, CASE WHEN <VoucherStatus> = 'STATUS2' THEN <ScheduledPayDate> ELSE NULL END AS ScheduledPayDate, CA

I’m working on an Outlook use case where we need to fetch meetings for a user within a specific start and end datetime using the Microsoft Graph API.ChallengeWhen calling the CalendarView endpoint, the Graph API requires both start and end datetime values. If no time offset is included, it defaults to UTC.For example, for Eastern Standard Time (EST), the offset should be UTC-5 hours.The question is: How can we dynamically pass the correct offset based on the user’s timezone?What we’ve triedWe attempted to format the datetime using the Moveworks DSL Format_Time function, but this requires the timezone in IANA format. From the Graph API, we only get the user’s timezone in Windows format (e.g., Eastern Standard Time). Is there a recommended way to convert Wind

If I setup a connection using authorization code, it will always make the actions based upon the initiating user. With this setup, I need the ability to have Moveworks make actions using someone else’s access. Here are some examples where this is needed:The initiating user requests for something that requires approval. A moveworks approval process is used to manage that approval. When the approver approves it, I need that to initiate an action in the same system but now using the approver’s account as the initiating user does not have access to approve their own request. With the new listener functionality, I see no way to run any action associated to a connection that has the authorization code grant setup as there is no initiating user.

Do we have the option to reuse plugins that we have developed in Agent Studio? We’ve created an approval plugin for one team within our organization, and now another team is requesting a similar solution. Although each team has different data points and systems they connect to, they both want the same approval flow process.I’m trying to determine if it’s possible to reuse the existing plugin to avoid hitting our plugin limit and to streamline development. If there’s a way to customize or clone the plugin for different data sources while maintaining the core process, that would be ideal.Could you confirm if this is feasible or suggest the best approach to efficiently handle multiple teams with similar workflow needs? Thanks!

We’re bringing together developers, innovators, and AI enthusiasts for a full-day hackathon at the iconic Chase Center in San Francisco on Tuesday, October 7, 2025. This one-of-a-kind opportunity to build right on the court puts you at the center of innovation where champions compete.✨ Special for Moveworks customers: Bring 2 or more developers from your org, and you’ll be entered for a chance to win a pair of the cutting-edge Meta AI Sunglasses*: smart glasses that bring AI right into your everyday life.Whether you’re building your first AI agent or pushi

I have a list of questions and answers in a restricted google document. It is in a sheet, but it can be moved to a doc as sheets do not seem to be supported. The use case is to ask the bot many different questions that are included in the document and have the bot return with the exact answer for each one of them. Currently this mostly works, but the answers are getting summarized. The answers need to be 100% exact as they are approved internally as correct. If I tell the bot to return the exact answer, it works most of the time, but is there a way to have it always return the exact answer instead of having to ask every time?

Is there a workaround for the output mapping result limit? I'm working on a flow where a user can query the bot to return all of their tickets in JIRA, but I'm losing some data in the process when the mapping gets bigger. Using conversational plugin, not compound action.Example Output Mappings:Mapping 1MAP(): converter: key: item.key project_name: item.fields.project.name status: item.fields.status.statusCategory.name summary: item.fields.summary url: $CONCAT(["https://company.atlassian.net/browse/", item.key]) items: response.issuesreturns:1. Project A – 38 tickets2. Project B – 1 ticket3. Project C – 1 ticket4. Project D – 13 tickets5. Project E – 32 ticketsMapping 2MAP(): converter: assignee: item.fields.assignee.displayN

Hi All,I am looking at removing a button presented in a notification to the user.Specifically I want to remove a button so the user only has two options. The button I want to remove from the notification shown is “Edit request” to avoid user confusion.Currently selecting “Yes” instructs the middleware to execute a remote action against the machine and clicking “Yes” again will fail as the remote action cannot be executed more than once on the middleware system.I cannot see anything in the CREST Event workspace that would refer to what looks like a consent option to proceed with the choice as shown above.