🎉 Agent Studio Office Hours – First Session of 2026!

Hey builders!

We kicked off the new year with a jam-packed session yesterday – myself, ​@rgeroulo , and ​@srankich  were cooking through new releases, live previews, and some deep dives into patterns that'll make your plugins way more reliable.

Here's the recap of what we tackled:

1. LDAP Actions – Finally Connect Agent Studio to Active Directory
   → No more building custom PowerShell wrappers! LDAP Actions dropped, letting you search users, retrieve attributes, and run any LDAP query directly from Agent Studio.
   → Requires an Active Directory connector in Moveworks Setup (not Agent Studio).
   → Docs: HTTP LDAP Actions
2. Data Mappers for Action Bodies – Goodbye Raw JSON Escaping Hell
   → New feature: use our familiar data mapper syntax instead of raw JSON payloads in your HTTP/LDAP actions.
   → Prevents LDAP injection – all dynamic inputs get escaped and sanitized automatically.
   → Works with any action type, not just LDAP. If you hate escaping special characters, this is your new best friend.
3. Optional Slots – Sneak Preview (Coming Soon!)
   → Steve showed off the upcoming Optional Slots feature with 3 new inference policies: Always Infer, Infer if Available, and Always Prompt.
   → You can now set fallback expressions using DSL – fall back to none, databank values, or action outputs.
   → Perfect for update flows where you want to keep existing values unless the user explicitly changes them.
4. Slot Collection Order – Left to Right Matters
   → If slot B depends on slot A, put A to the left of B. The leftmost slot gets collected first.
   → Not super intuitive in the UI yet, but it's how you handle dependent resolver strategies.
5. Displaying Data Mid-Conversation – Output Mappers, Not Content Activities
   → Use display_instructions in your output mapper to steer how data is presented – but keep it simple (bolding, ordering).
   → Don't cram data values into display_instructions! It's non-deterministic. Use the mapper to structure your data, use instructions just to steer.
   → For lists, always use MAP() to iterate properly.
6. Structured Data Analysis (SDA) – Big Data, No Problem
   → SDA auto-triggers when your payload is large. It converts natural language to code at runtime to filter/sort/analyze.
   → Handles up to ~4MB of structured data.
   → Check the SDA Quickstart and Data Retrieval Cookbook.

Here's the recording to revisit any part of the session or share it with your team: Recording link

Resources to Bookmark

→ The Agent Builder Newsletter – AI news and Agent Studio insights, dropping weekly → Data Retrieval Cookbook – patterns for handling slots, resolvers, pagination, fuzzy matching → Agent Architect – your copilot for building plugins

Catch You Next Session

→ Check upcoming sessions and register here: Developer Office Hours

See you there 🤖