Changelog

Hello everyone, a lot has changed over the past two months, from new nodes and automation to team features. Here is what's new in Upstage Studio.

1. New Validate node: check extracted results against your rules

You no longer have to review every extracted value by eye. Set rules such as "the total must equal the sum of the line items", and the Validate node marks each result green, yellow, or red and routes it to the next step based on that verdict.

Describe a rule in plain language and Studio builds the condition for you, so there is no need to write complex formulas.

2. New Merge node: bring split results back together

When one file contains several kinds of documents, Studio splits them by type and extracts each one separately. The Merge node combines those separate results into one and passes it on to the next step.

You can summarize the merged result with Instruct in a single pass, or use Validate to compare values across different documents.

3. Triggers: run automatically when a new file or email arrives (Beta)

You no longer need to upload files by hand every time. Set up a trigger on an agent, and it picks up new files in your connected Google Drive or OneDrive and new emails in Gmail or Outlook, then runs automatically.

You can review what was processed on the trigger screen, and the agent list shows at a glance which agents have automation connected.

4. Document translation in Grace

Attach a document to Grace and ask "translate this into English", and you get a translated version that keeps the original layout and images as closely as possible. Grace first prepares translation guidelines for the document, then translates and runs a quality review. When it finishes, you can compare the original and the translation side by side.

Try it with a sample document from the Translate button on the home screen.

5. Grace chat does more

Grace now helps with much more than building agents.

• Document Q&A: Ask questions about an attached document and get answers with sources. Click a source to jump to that spot in the document.

• Web search: Grace finds the information you need on the web and answers with sources.

• Connected apps: Grace can use your connected apps, such as Slack, during a conversation. For example, it can send results straight to a Slack channel.

• Model selection and Canvas: Choose the answer model from the input box, and preview code or HTML output in the Canvas next to the chat.

• Approval mode: Choose whether Grace asks every time before actions that cannot be undone, such as sending a message, or decides on its own.

Chats are also titled automatically from your first message, so past conversations are easier to find.

6. Organization and Projects for teams

You can now use Studio as a team. Create an organization, invite members, and manage agents and members by project. Agents can be moved to another project while keeping their run history.

Organization settings bring API keys, usage, and audit logs together in one place. Organization invitations and project notifications can be handled right from the sidebar notifications.

7. New pricing: organization subscriptions

As of September 21, Studio pricing has moved to organization subscriptions. Choose from Free, Small, Large, or Enterprise, and subscribe to as many seats as you have members.

On the Billing page in settings, you can review the amount due and the effective date before changing your plan or seats, and the Usage page shows how much usage you have left. With this change, the 10 free runs per agent have ended.

See the pricing page for details.

8. Node settings and result downloads

• Parse: Set the page range and image quality to process in advanced settings.

• Classify: Move a sub-category up one level in hierarchical classification, and see the number of classes on the schema card.

• Instruct: A button that writes the prompt for you and a web search option have been added.

• Extract: Choose Download all to get the extraction results for many documents at once. The download includes both the results with your corrections applied and the original extraction results.

9. Easier navigation and management

• Pin the agents and projects you use most to the sidebar as favorites, and collapse or expand sidebar sections.

• Resize columns and choose which columns to show in the agent list and other main tables.

• When you delete a chat or an agent, choose which uploaded files to delete along with it.

• JSON file upload is supported, and you can download the original file right from the document viewer.

• The term "file group" is now "case" across the product.

10. Library and Monitoring

The agent detail page in the Library is now easy to read on a phone. A selector at the top of the results lets you switch between step results, and public Library pages can be embedded in external websites.

In Monitoring, you can see why a failed job failed, right in the job list and the detail view.

Alongside all of this, we've shipped many stability, performance, and security improvements.

As always, we'd love to hear from you. Leave your feedback and feature requests anytime.

Thank you for building with Upstage Studio!

Hello everyone

A lot has shipped over the past few weeks — here's what's new in Upstage Studio.

1. Import files from Google Drive, Gmail, Outlook, and OneDrive

You no longer need to download files to your computer before uploading them again. Connect Google Drive, OneDrive, Gmail, or Outlook and bring documents straight into Studio. For email, you can browse your Inbox and Sent folders and pick attachments right under each email — or even import the email body itself — which is especially handy for invoices and orders that arrive by mail.

2. Upload folders — and more file formats

Drag and drop an entire folder and every document inside uploads in one go. Upload limits have been raised and unified (up to 500MB / 1,000 pages per file), and Studio now supports more formats, including MSG, EML, MHT, HTML, TXT, and MD.

3. Finer control over runs

Managing agent runs is now easier from start to finish. You can run only the documents you've selected instead of the whole set. If a node's settings are invalid, the Run button tells you exactly what to fix before anything starts, instead of failing mid-run. Jobs in progress can be canceled right from the Run button — multiple at once if needed. And when a run finishes, a completion notification and a "new results" indicator appear, so you can work on something else and come back when it's ready.

4. A clearer editing experience: drafts, validation, and run notifications

The editor now does a much better job of telling you what's going on.

• Unsaved changes (drafts) are clearly marked, and discarding them shows a confirmation with an Undo option — no more accidentally losing work.

• If a node's settings are invalid, the Run button tells you exactly what to fix before you run, instead of failing mid-way.

• When a run completes, a notification and a "new results" indicator appear, so you can return when the results are ready.

5. Schema editing quality-of-life

Small things that add up: reorder Classify schema keys with drag and drop (the same UX as Extract), resize schema editor columns, and use Cmd+Z / Ctrl+Z to undo and redo while editing configs.

6. Hierarchical classification in Document Classify

Document Classify now supports nested categories. Instead of one flat list, you can organize document types into a hierarchy (e.g., Invoice → Domestic / International) — much easier to build and maintain as your classification scheme grows.

7. A more powerful chat

You can now run your agent directly from the chat, and clicking a result chip jumps straight to that spot in the results with a highlight. You can keep typing while a response is being generated, and stop generation anytime with the stop button. File handling is faster too: multiple files upload in parallel, you can send your message while uploads are still in progress, and already-uploaded files can be reused with other agents without uploading them again. A new-conversation button in the sidebar starts a fresh chat in one click.

8. A brand-new Library experience

Library agent pages have been redesigned from the ground up. Each agent now shows real sample documents — click one and watch the workflow run step by step, with the output of every stage. Found an agent you like? Hit "Try with my documents" to copy it and upload your own files right away. You can also browse the Library by category from the home screen.

9. Easier agent sharing and portability

Sharing configurations is simpler: share all of an agent's configs at once, or download them in bulk to move work between environments. Shared agents can now be deleted in one step, with an automatic unshare.

10. Monitoring upgrades

The monitoring page now breaks down metrics per workflow step, surfaces failed runs so nothing slips through unnoticed, and lets you download your job list as an Excel file for reporting and further analysis.

Alongside all of this, we've shipped a large batch of stability, performance, and security improvements.

As always, we'd love to hear from you — leave your feedback and feature requests anytime. Thank you for building with Upstage Studio!

1. Edit existing agents with Chat Generator

Chat Generator now supports not only creating new agents, but also modifying agents you’ve already created.

Previously, Grace focused on helping users create agents in the editor. Now, it can also support the next steps after creation, including schema refinement, configuration changes, and agent restructuring through chat.

We’ve also added the ability to revert to a previous Config, making it easier to manage changes and iterate more flexibly.

Instead of adjusting complex settings one by one, you can now improve your agent by simply chatting with Chat Generator.

2. Auto-run option added for uploaded files

We’ve reduced the extra step of manually running a workflow after each file upload.

When Auto-run is enabled, the workflow starts automatically after a file is uploaded. This makes it faster and smoother to use agents when you need to upload documents repeatedly and check the results.

3. Apply settings to multiple Extract nodes at once

We’ve improved the experience of updating settings across multiple Extract nodes.

Now, after changing the settings in one Extract node, you can select Apply to all nodes to apply the same settings to other Extract nodes at once.

This makes workflows more efficient when working with multiple document types or repeatedly adjusting model settings during testing.

4. Copy and download results more easily

We’ve added copy and download buttons so you can use node results more easily.

You no longer need to manually drag and copy results. With a single click, you can now copy or download outputs from Instruct, Classify, and Extract nodes.

This makes it easier to document results or move them into other tools for further use.