Agent Studio

You can now collect files from users right inside Moveworks! With File Slots, you can design plugins that request and receive files — like receipts, documents, and invoices — and send them directly to external systems.  What's included:• File uploads via a built-in "Upload File" button in chat• 🎕 New: Multiple file uploads now supported via List[File]• Seamless integration with HTTP Actions through flexible file mapping• Supports raw upload, multipart/form, multipart/related, application/octet-stream, and base64 encodingSupported file types: .png, .jpg, .jpeg, .pdf, .txt, .csv, .xlsx, .xlsm, .xls, .xltx, .xltm (up to 5MB each)📖 Learn more:• File Slots docs:

I’m facing an issue with a Compound Action where the flow executes successfully until fetching the manager_record, but the notify step does not trigger. There are no errors in the logs, both employee_record and manager_record are not NULL, and the workflow does not explicitly fail or terminate. The execution simply stops after fetching the manager record, and I don’t see any logs beyond that point. For testing, I’m using a single static record via script, but in the actual use case, the data comes from an HTTP action, which is why a for loop is used. I’m also attaching a screenshot of the logs for reference.Here is the exact Compound Action configuration:steps: - script: code: | return [{ "LOGON_NAME": "mw_test_user_5@company.com", "EMAIL_ADDRESS": "mw_test_user_5@company.com", "MANAGER": "manager-name", "ACCOUNT_EXPIRY_DATE": "2026-02-26 0

Hey everyone, first time posting here, I’ve tried out the User Consent Auth / Authorization Code Grant between Moveworks and Workday, functionally it is working well after user consented, but is there anyway to detect/test if a user consent expired/non-exists at the beginning of a plugin?Problem StatementIf a user never consent previously or if token already expired, the first time calling a plugin will always fail (not yet consented / token expired). Immediately afterwards, User Consent Request will appear: User can then proceed to consent, but the plugin already ended so user will have to manually retrigger the plugin after consented, and retype everything again. QuestionI need to make my plugin:1) Check if user consent exist/valid2) If false in 1) then show User

Since from two days many users are reporting a popup issue, there is no change made from our end like manifest file for anything w.r.t our bot. its not specific to certain skill, happening for all. when yo cancel it or “Not now”this shows “this card action is disabled because of you blocked the bot” anybody facing similar issue?  

Hi,Is there any workaround to catalog item form fillability for ServiceNow catalog item forms.most of our form are still not fillability due to client scripts.Regards,Rajeev

We had our highest attendance yet, 40 people on the call and the questions did not let up! Ryan was meeting with a customer and Steve was still waking up on the West Coast, so I ran this one solo — and honestly, it turned into one of the most teaching-heavy sessions we've done. I spent the first chunk breaking down how LLM reasoners works under the hood, then we got into live debugging, scoping gotchas, and a sneak peek at the observation interpreter.If you've ever wondered _why_ your plugin works sometimes and falls apart other times, this session was for you. Let's get into it.How the Reasoner Works: Context, Attention & Why Less Is More This was the big one. I walked through how the LLM actually processes everything in Agent Studio — your plugin config, slot descriptions, conversation messages, action outputs — all of it gets packed into a context window as tokens. The reasoner uses attention to figure out what matters, but here's the

Hi community,Would it be possible to reconsider deprecating the Events section in the Moveworks Classic Agent Studio? At the moment, it’s no longer possible to create new projects using this functionality.In my view, the Events API and listeners serve different purposes and are not interchangeable.The Events API is well suited for simple automation use cases where there are no dynamic variables and minimal complexity - such as sending predefined, global messages. It’s lightweight, easy to configure, and very effective for these scenarios.Listeners, on the other hand, introduce additional features which is perfect if we want to show buttons and do event processing, but also complexity and they have timeout limitations associated with compound actions.For these reasons, retaining the Events functionality would still provide value for straightforward use cases.Thank you for considering this.

Hi everyone,I am building a Plugin in Agent Studio that retrieves a user's timecard details and presents them with an "Approve" button using a Compound Action. I am running into an issue where the Reasoning Engine hallucinates a successful approval status immediately upon displaying the card, rather than waiting for the user to click the button.The GoalPlugin triggers (e.g., "I need to approve my timecard").Compound Action fetches data, formats it using LLM, and sends a notify message with an "Approve Timecard" button.The Bot should say: "Here are your details, please review and approve."The IssueWhen the plugin runs, the bot outputs the following message at the top of the response sometimes at the end, before the timecard details:"Your timecard has been retrieved and successfully approved."It seems the AI believes that simply running the plugin succ

Hey builders!Steve, Ryan, and I were in the kitchen today – we launched Optional Slots live on the call, went deep on passing data between actions, tackled output formatting best practices, and walked through user consent auth with Microsoft Graph.Here's the recap of what we crushed: Optional Slots – Launched Live on This Call 🔥 → Three slot inference policies now available: Always Ask, Infer if Available, and Always Infer. Always Infer means the assistant never prompts for that slot – it falls back to a DSL expression or null. → Two big patterns: smart defaults (fall back to a recently fetched value) and null values (let unused fields pass as null for generic update plugins). → Docs: Optional Slots

Hi team, I am building a workflow using the notify action with a webhook listener. I want to send a message to the end user that is preceded by a greeting. When using the classic events, there was an option to use a greeting (Hi, {{name}}). Is this something that we can turn on with the new webhook listeners? If not, what would be the best way to implement this? 

Sharing Solution to HTML-encoded string passed from Webhook call to present URLs to users through Teams Notification (includes mailto: clickable link and URL) Notify action to handle link:steps: - action: action_name: mw.get_user_by_email output_key: mw_get_user_by_email_result input_args: user_email: data.email - notify: output_key: output_notification recipient_id: data.mw_get_user_by_email_result.user.id message: RENDER(): template: "Hello {{name}}{{& message}}" args: name: data.mw_get_user_by_email_result.user.first_name message: data.message.$REPLACE("&", "&").$REPLACE(""", "\"").$REPLACE("<", "<").$REPLACE(">", ">")

I am leveraging optional slots in a conversation process I have been building, but have 2 issues:The conversation failed on setting a slot called conversation_context, that is set to Always Infer, with this error “hit an internal issue trying to build the conversation context and that specific run (execution <execution id>) has been stopped and cannot continue”. This has only happened once so far. I have a user object slot that is getting populated via a notify statement. I have tried all 3 infer options and none of them have been successful in setting the user properly. I can no longer get this functionality to work that was working perfectly fine earlier today.This last issue is big in that it seems to be is breaking notify statements sending a user object to a conversational process.

I have a slot called description where I am asking the user to describe what they are looking for. In general this just ends up being what initiated the plugin. I want to enforce that they provide enough information for it to be handled properly. I see 2 ways to do this currently:Put in a slot validation based upon the overall length, number of spaces, etc Have a second string slot to ask follow up questions#1 does not reliably work as moveworks just adds in additional words to make it long enough 90% of the timeI would prefer not to do #2, as it is possible the user could be providing enough information so additional prompting may not be necessary.Is my best bet just having a secondary slot or does anyone have good ideas to handle this reliably?

Introducing Optional Slots in Agent StudioWe're excited to share a new capability in Agent Studio that gives you much more control over how your conversational plugins collect information: Optional Slots.The problemUntil now, every slot declared in an Activity or Decision Policy required a value before the plugin could move forward. That meant users were prompted for every single field if no value was explicitly provided when NULL, a previous step, or a sensible default would do. For plugins like quick updates, lookups, or one-shot transactions, this added unnecessary friction to the conversation.What's newOptional Slots introduce two key concepts: More robust Slot Inference Policies and Fallback Values.Slot Inference Policies let you control how each slot's value gets collected:Always Ask: The assistant always prompts the user, even if a v

Hi Moveworks Community — we’re looking for help troubleshooting routing/trigger conflicts and candidate list selection behavior for an Agent Studio software lookup experience. I’m including sanitized configuration details below (no internal URLs, no people info).1) What we built (use case + goal)We built an Agent Studio plugin to let users search an approved software SharePoint list by software title, including spacing/variants and misspellings (e.g., “Power BI” vs “PowerBI”). Output should show Title + Approval Status + (Restrictions only when applicable) + Corning Standard + Order/Access URL.2) What Moveworks engineering has already done (important context)We worked directly with a Moveworks engineer and have already implemented multiple changes/guardrails:Added steerability / deprioritization for forms to reduce surfacing of software request forms in th

If you're exploring Agent Studio, you'll come across plugins. These aren't your typical software components. Think of them as skills you can teach your AI agents to do more, move faster, and scale across your enterprise.Our Staff Developer Advocate ​@Kevin Mok breaks it down in under a minute 👇Video linkWant to go deeper? Join Kevin for developer office hours this Thursday at 1 PM PST: Register here.

I have a compound action that is around 100 lines long but is only 6 actions. I cannot publish or test it as I continue to get the following error:Message: Sorry we cannot process your request this time.I have already broke this set of actions out from another compound action as I wasn’t able to save that one until doing so. I believe this has to do with the size of my compound action as a while back I saw a 5mb size limitation error message when running into this error in another situation. Is my only course of action to break this compound action out into further smaller compound actions?

Hi Team,I have created content gateway having Authentication type selected as API Key. The ingestion is failing with error as Validation Failed. I assume it is due to the header is not getting passed in the request because when I am sending the request in API Playground it asked me for the Header Key and Value.While configuring the Content Gateway connector there is no option to add the header as we have in API Playground or HTTP action , how we will handle this ? Also is there a way we can see the ingestion logs ?

I am building an extend expiration date for contractor skill in agent studio, agent architect really helped me a lot. this skill I already developed on creator studio (legacy). now building in agent studioLooks like Data type with resolver strategy is not kicking of.. backend logs i dont even see the API being called because of this i am getting error in data.contractors not having the data to perform for each.. (check snapshot)how do i troubleshoot?  I do have a 2 http one to get contractors for the user so among that only user can extend it.  API: get_manager_reportees_actioninput: email{    "result": [        {            "u_termination_date": "2026-03-17",            "u_commonname": "Joe Doe",            "email": "Joe.Doe@abc.com"        }    ]}another to update the dates PowerAutomate_update_user_expiration_action Created a data

I am creating a plugin that sends a Slack message to a user when triggered by a webhook from a third party system when the system is experiencing service outages. The plugin is configured with a webhook trigger and uses a compound action in its body. The compound action has a user_email input argument to find the recipient to send the Slack message to, along with a message template that parses some of the information from the webhook’s payload. In my testing, I am sending a mock webhook payload via curl command and providing an email address of a user to send the Slack message to, but the actual payload from the third party system does not include email addresses/user information. Instead of sending the Slack message to just a single user, I’d like to send it to either all Slack users (which Moveworks has ingested), or a group of users, for example that are identified through an email distribution list that use the third party system.Is there a way to accomplish this so the p

Hey builders!Ryan and I had a packed session today – we went deep on data mappers in HTTP actions, tackled some gnarly edge cases around webhook security and output mapping, and dropped some roadmap previews for what's coming in the next few months.Here's the recap of what we covered:Data Mappers in HTTP Action Bodies – Your New Best Friend for Transformations → You can now use YAML-based data mappers instead of raw JSON with mustache syntax in your HTTP action bodies. → The real win: use DSL functions directly – uppercase, map, inline transformations – without wrestling with triple braces or escaping JSON strings. → Perfect for abstracting transformation logic out of your conversational process. Ryan called out Outlook meeting booking as a prime use case – get a list of attendees, pass it to the action, let the action body handle the restructuring. → Docs:

Encountering an issue when using the Notify action inside a for-loop (Under parallel) in a compound action flow. The Notify action is configured with two interactive buttons: Approve and Reject, and is used to send actionable notifications to users.When the loop processes a larger list (10+ items), users do not receive any notifications. Instead, the conversational flow appears to stall, and the user continuously sees the response:“Fetching invoice details for your pending approvals…”In contrast, when the list contains a smaller number of items (typically fewer than 5), the Notify action works as expected and notifications are successfully delivered to the user with the approve/reject options.There are no errors or failures visible in the logs during execution. The workflow does not explicitly fai

We were updating the names for our plugins to be more user friendly, however when removing the Underscores we received an error. Once I read the error explanation it said, ‘Name should start with a letter and underscores are OPTIONAL.’So why is this error happening? How can I save without underscores?  

We are working on a use case , where we get invoices from databricks , with requesters username, associated with that . what our current flow does is that , it fetches username for each invoice gets its email and fetches the profile of the user to get the email and region of the user and also to fetch the manager information to send the invoice escalation notification if an invoice hits deadline on that particular day. So the logic works in a way that it runs daily , and groups invoices based on their ages and as per the requester and managers and sends those to the respective people. This is flow is working successfully . But theres is a catch . On Biweekly Wednesdays we have the logic to pull all invoices above a certain number of days. So it fetches thousands of invoices on Biweekly wednesdays and then tries to process them. But while processing the script gives an error saying “ Deadline Exceeded” with the error code 6. As we are not calling any external API why is the code fail