Agent Studio

Hi everyone,I am looking for input on how other teams handle plugin testing and lifecycle management. I'd appreciate hearing what is working well for you!Our current process:We build and test in our Sandbox environment. We temporarily add testers, and once testing passes, we remove them to keep the environment clean and restrict access. We then export the plugin to Production and run a quick smoke test to confirm it works. For enhancements: We build the changes in Sandbox, export the new version to Production, and then swap them (turn off the old Production plugin and turn on the updated one). We have to keep the original plugin active in Production while we build the enhancement in Sandbox. What we are trying to figure out:We are looking for a simpler approach. We considered just building directly in Production and using the Launch Configuration to whitelist our testers (via "Allow selected users"), but we hit two roadblocks:Syste

I am working on a ambient agent that is able to send notifications to 1 or more users. The listener is expecting a payload with an object that has email and the message to send to the user. I wanted to know if there is a size limitation on the API call coming into the listener. Trying to determine how much to limit the bulk api call size if needed. 

My compound action sends two visible messages to the user: an intermediate "Cancelling your request..." message that appears while the action is executing, then the final "has been cancelled" confirmation after the action returns. I want only the final message to show.I've tried strengthening display_instructions_for_model on the return mapper to explicitly suppress intermediate narration ("Do NOT generate any in-progress messages, output ONLY the final confirmation"), but the duplicate still appears.Is there a plugin-level, CA-level, or workspace-level setting that controls whether intermediate progress messages are streamed? If not, is there a documented workaround?Platform: Agent StudioChannel: Teams 

When I used the provided cURL in my HTTP action to retrieve data and connected my connector, the test executed successfully. However, the file returned is empty.curl -X GET "https://marketplace.moveworks.com/api/purple-suite/erp/purchase-requests/history"Am I setting this up incorrectly, or could there be an issue with the API itself?

Building a Cancel PTO plugin in Moveworks against Workday by using built-in Workday Cancel PTO Plugin. Hit a snag detecting the third state of a time-off request.Setup:Workday_GET_Approved_PTOs_Of_Employee_ByDate calls timeOffByDateTaken and filters on status = Approved WIDAfter an employee initiates cancellation via Adjust_Time_Off_Request, the entry shows in Workday UI as "Submitted (Cancelation)" awaiting manager approvalBut WQL's status field on these entries still returns "Approved" — so the plugin can't tell they're already mid-cancellation, and users could trigger duplicate cancellation requestsQuestion: Which WQL field (or wh

I built a Cancel PTO plugin using the Workday Correct Time Off API. Workday side works correctly — BP routes to the manager, approval task shows pending with the right awaiting person, and our custom approval report picks up the correction record.The issue: the approval card never reaches the manager in the bot. Request Time Off approvals work fine through the same pipeline; only corrections don't push through.In Approval Configuration → Business Process Mappings, the Default Resource Type dropdown only shows Request Time Off and Leave of Absence. No Correct Time Off option, and no visible way to add one.Questions for the community:Has anyone enabled approval notifications for Correct Time Off / Cancel PTO in DOT? How do you add a new resource type or BP mapping for approvals — UI, support request, or backend config? Should approvals for a custom Cancel plugin ride on

I am using Try-Catch-Raise inside Compound Action to handle error Status Code scenarios and show user static message.Example:Status Code 401 = return static message with a link to request access.Status Code 500 = return static message with a link to raise incident ticket.But for Status Code 400, I’ll need to return different message to user depends on the response error JSON object.The Raise step has an output_key which can store the error JSON object:  May I know where and how can I access+use this output_key?I tried adding Action/Return/Switch after the Raise, but none of them works because plugin will end gracefully after Raise.If I change the Raise to Return (so that it is now Try-Catch-Return, no more Raise), how can I map the response’s error object in the Return?A

Hi Moveworks Community,I’m currently working on configuring Ivanti IT Knowledge articles ingestion within Moveworks via the following path:Moveworks Setup → Enterprise Search → Search → Configure Search → Classic Ingestion → Internal KnowledgeI wanted to understand how we can implement user-specific permissions or access control for knowledge articles ingested from Ivanti IT.Specifically: How can we ensure that only authorized users or user groups can access certain knowledge articles? Does Moveworks respect source-level permissions from Ivanti IT during ingestion, or do we need to configure access control separately within Moveworks? Are there recommended best practices for mapping user roles/groups between Ivanti and Moveworks to enforce these restrictions? Any guidance, documentation references, or examples would be really helpful. 

If you just walked out of the Moveworks lab or CreatorCon expo at Knowledge 2026, you built a real plugin. Triggers, a conversational process, slots that resolve from natural language. The same primitive every Agent Studio plugin in production uses.You have the muscle memory now. The next move is using it on something that matters to your team.Come to the next office hours.Live, builder to builder, no slides. Bring whatever you're stuck on, or just listen in. We’ll be hosting sessions on:May 21 June 2 June 18Sign up hereDrop a comment on this post, get a CreatorCon community badge (you’ll need to be signed in to do this).Tell us:A use case you want to build, or already built A takeaway you're bringing back to your team

I have a webhook trigger configured in Agent Studio that is connected to a GitHub Enterprise webhook listener. The webhook receives pull_request events from GitHub and is intended to only trigger the plugin when the action is review_requested or ready_for_review. To enforce this, I have set the following event filter on the webhook trigger: parsed_body.action IN ["review_requested", "ready_for_review"]. However, the plugin appears to be triggering on all pull_request actions sent by GitHub, including closed, regardless of the event filter.I have verified that the filter is saved and the plugin is in a published state, and I have also tried alternative syntax using parsed_body.action == "review_requested" OR parsed_body.action == "ready_for_review" with the same result. The payload arriving from GitHub is correctly structured with the action field at the root level of the bo

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

ProblemA common challenge in Agent Studio is how to serve grounded answers from public-facing documentation without building a full external ingestion pipeline.Teams often want the bot to answer questions using content that already lives on public websites, but they also want the experience to staygrounded in source material rather than general model knowledge transparent, with citations or traceability lightweight to implement and maintain targeted enough to return useful answers without introducing too much ambiguityThis becomes especially relevant when the source content is public documentation, such as help centers, product docs, compliance pages, or accessibility documentation.A good example is accessibility content. A user may ask about a specific capability such as Teams closed captioning, and the desired behavior is for the bot to answer from the correct Microsoft acc

I am attempting to create HTTP actions to support my plugins, but I am receiving an error message when validating the action. I am getting an Error Code 4, with Trace ID _Nenopmb-cr8f16u.I have attached a screenshotof the error message as well. Please reach out if you have seen this error, or know any troubleshooting steps. Thanks!  

I wanted to share a practical Agent Studio plugin idea we’ve had success with: using a scheduled ambient agent to scan recent support tickets and proactively notify responders when an outage pattern emerges.The problem this solvesIn many environments, the first signal of an outage isn’t monitoring — it’s a sudden spike of tickets describing the same issue.Example: within an hour, several users report symptoms like:“VPN won’t connect” “SSO login failing” “VDI keeps disconnecting” “Email delayed / cannot send”If responders only notice once tickets are manually triaged, you lose valuable time. The goal here is to detect the pattern automatically and alert a targeted responder group with a crisp summary and relevant ticket IDs.High-level approachSchedule: Every 30 minutesLookback window: Ti

I have a plugin for ticket search. It collects a search description and uses a LLM to collect/format optional attributes that are used in the search api. I am paginating the api up to 5 times in a compound action and don’t want to enable more than that. This works with simple search queries in that Structured Data Analysis can format the results as requested. However if I ask for a medium to large data set (over 500 tickets), it seems to go into a infinite loop showing 8-20 messages to the user saying it is searching for the data until it finally seems to time out and break without showing any data. I don’t want the plugin to loop especially as it doesn’t seem to work. I tried asking Agent Architect, but its suggestions didn’t help. How do I solve this problem?

I am trying to use Agent Architect but since morning it is returning 400, 429 error codes.I have been using it frequently and has been of great help but today it is not working at all.Hence wanted to check if it is down ?I have retried several times throughout the day but everytime it shows error  

Newbie customer hereI want to add a filter that will exclude asking for approval for change tickets that are CAB Ready TrueI found this command and tried to add it to Universal Ticket fetching filterNOT (ticket.type = "change" AND ticket.cab_review = true)but it did not like thatwhere do I put it?  

Hello Hello All,  We are getting our feet wet with Moveworks and we are loving it so far. We have a prod and a sandbox environment and we know they are separate. Our prod bot in teams is associated with the OOB teams bot and for the sandbox, we created a non-app store and we have it already uploaded into our Teams DEV tenant. It does not seem to work at all.. it does not respond.. Anyone here tried that same approach that can share the challenged encountered and how to overcome them?

Hello Community!One of the fastest ways to deliver some productivity use cases in Agent Studio is to pair a simple slot with the out-of-the-box generate_text_action. In a single plugin (or multiple plugins for more granular analytics), you can spin up productivity use cases like code generation, email drafting, content generation, brainstorming, or rewriting content — no custom LLM integration required, no API keys to manage, no prompt-engineering framework to stand up.This post walks through the architecture we use to build these plugins quickly.The Solution: slot → generate_text_actionThe pattern is intentionally minimal. You collect the user's request through a slot, hand it to mw.generate_text_action with a purpose-built system_prompt, and

Hi Team, How to add new users in move works to get them access to move work instance (our company owned Move works instance)? is there any documents or kb articles that supports from query please share with me. ThanksUpendra T

Hi All, How can I configure the “New Hire” functionality is employee comms.At the moment, our new users are being identified from ServiceNow user records, the condition will look like followingcreated : today user type is not functional OR externalRegards,Rajeev  

Managing all the connectors I have in Moveworks Setup is a lot and I want to be able to easily track when an API token is expiring for Moveworks in Atlassian for example to avoid degradation in our Moveworks Assistant due to expired credentials. How can I use Agent Studio to notify me 30, 15, 5, etc. days in advance before expiry? Is this possible?