Agent Studio

In the Hackathon, the roadmap of Moveworks was presented to us. We were told that voice integration is in the pipeline. Do we have the timeline? Where can I find more updates on the progress? 

I have a api call to get a list of categories. From there I have another to get a list of subcategories. I see slot resolver strategies do not have access to the data element from the conversational process. Is there a way I can pass this information with input variables in a reliable way without asking the user for the category twice and letting them select a improper value?

I have tried both removing the progress updates as well as customizing the progress updates and I have had no luck with either. Is this a Moveworks bug or is there some other way to disable these progress updates apart from not including them in a compound action? Here is an example of the behavior I want to avoid.  

🎉🚀 Agent Architect Visualization Upgrades!!! 🚀🎉We just came off a month of intense hacking with all of you at our San Francisco, London, & Bangalore Hackathons. First of all, THANK YOU so much for coming out – it was very inspiring seeing everything you built & continue to create. We heard from you at our hackathons it’s frustrating when Agent Architect generates Compound Actions or Data Mappers in JSON instead of YAML, NO MORE!Agent Architect code outputs can now be toggled between YAML & JSON with a click. BONUS – Agent Architect diagrams (Mermaid) can now be visualized, expanded, & downloaded 👉 Tell me what you wanna see next! Happy hacking!- Ajay & The Agen

i have an agent studio plugin that uses the User object as a slot to get a user context. what i’m finding is the bot is responding with a default set of properties even though i’m not specifying which values from the user object to display. i’m using the built-in mw.generate_text_action and defining a slot in the conversation process as a User object type, which is grabbing the user context correctly. but the bot response message is showing the same default properties (full name, email, role, manager, dept, location) and i don’t want to display those, at least not like that. i tried using display_instructions property in the output mapping, but that doesn’t work. the default properties are still displayed.where is the plugin getting this format from? and how do i override it or define my own template?

Any Idea why my user_email variable would not be passed through here or be blank as you can see in the canonical_endpoint? I have also tried replacing “user.email_addr” with “mw.user.email_addr”. Why are neither of these working? Is my syntax incorrect? Or could I have incorrect bot configurations? I wouldn’t need a slot here would I? 

disregard… i’ve simplified the plugin and took out the compound action and just referenced the built-in action in the conversational process directly. it’s sending one message now, but it’s not the right format. that’s a separate issue though, so i’ll create a separate post for that. this one can be closed.i have a plugin with a conversational process that runs a compound action. the compound action runs a built-in action that returns a single value.the bot is responding with three messages, which is what i don’t want. where are those messages coming from? my understanding was the bot will send a message AFTER the last action completes. if there is only one action, why is the bot responding with multiple messages?the last message sent is more or less what i’d want displayed to the user, not the others. 

We have on prem agents connected to our ITSM (ME/SDP).  Any way to use them via bot to reach into an end users’ machine to add them as a local admin to conduct elevated tasks - but them remove it after a time limit?  Thanks!

I am able to get a JSON object root response to work, like this: { "Items": [ { "name": "root?", "thing2": "asdf" }, { "name": "root 2?", "thing2": "asasdfasddf" } ]}With Mapper like this:output: MAP(): items: Items converter: name_output: item.nameOutput is this (as expected):{ "output": [ { "name_output": "root?" }, { "name_output": "root 2?" } ]}  Problem is when the input (response object) is an array of objects with no root key to address:[ { "name": "root?", "thing2": "asdf"

My team has recenlty started testing around with installing Agent templates from the Marketplace into our TEST environment. Once particular agent template is Summarize an Incident. When testing this we identified a potential issue where any associate could get a Incident summary for any Incident. I see this as a potential security concern as I wouldn’t want a non ITIL user to be able to see higher severity Incidents as an example. Has anyone else implemented this Agent template? If so, have you had similar experiences or have any idea how to implement the SNOW Role-Based permission mirroring?

I’ve setup a Rapid Global notification when a user signs in or out with Webhooks.The challenge I have is I don’t want to enable the notification for everyone, however, Webhooks doesn’t allow you to select an audience to restrict, the Audience option is disabled.It would be useful to be able to control the notification audience on the Moveworks end rather than have to adjust 100s of users in Rapid Global.  

I have an ambient use case, that is triggered when someone books a meeting room in Outlook. The webhook and compound action all work ok, and trigger my conversation process.However, my conversation process starts off with a long introduction about our preferred catering suppliers, that I needed to be verbatim text, so I placed it into an action that is called at the start of the process. No slots are required and I tell it to wait for the action to finish before proceeding: Only in the next action do I ask it to collect the Catering Site, which is a list of options in a resolver strategy:However, when the conversation runs it usually (not always), shows my long blurb after the

I am having issues setting up a Oauth2 for Microsoft products. I cannot get a successful API call even though I am able to with client credentials in Postman. Have others been having issues with this? Is there something I’m missing. I’ve been following this guide amongst trying many other things: Connector | Microsoft Graph

Hi Team, For the update feature requests I was trying to use compound action rather than the HTTP actions in the COnversation process. I created the compound action but it is not getting triggered . Could you please have a look into the compund action and let me know if in case there is something I am missing steps:  - action:      output_key: Mw_Get_Requests_result      action_name: Mw_Get_Requests      progress_updates:        on_complete: ON_COMPLETE_MESSAGE        on_pending: ON_PENDING_MESSAGE        message: "{{Mw_Get_Requests_result}}"   - switch:      cases:        - condition: feature_request_id == 'FR-50911'          steps:            - action:                output_key: submitted_requests                action_name: Mw_Update_Requests_EXAMPLE                progress_updates:                  on_complete: ON_COMPLETE_MESSAGE

We’re still buzzing from the Moveworks AI Agent Hackathon, where the Golden State Warriors court transformed into a playground for agentic AI innovation.Hundreds of developers, customers, partners, and AI innovators came together at the iconic Chase Center, to build the future of agentic AI (and win some exciting prizes too). The energy, the creativity, and the collaboration were second-to-none.Attendees heard from Moveworks product leaders, customers, and partners on the latest innovations shaping our platform, including the power of Ambient agents that act across systems on real-world challenges including:Resolving lost package resolution that goes from frustration to fix in record time Managing CRM updates that keep sellers selling, not entering data Enabling expense compliance that gives finan

 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

Is it possible to reference and manipulate a user attribute within a Content Activity? For example: grab a user’s email username@email.com and remove the @email.com so that it returns only username