What is Studio?

Upstage Studio is a no-code platform for building, running, and improving document agents — without writing code.

Instead of handing an entire document to a single model and hoping for the best, Studio splits the work into AI nodes with clearly defined jobs and connects them into a pipeline. You can inspect and correct the output of every stage, which is exactly where consistency and verifiability come from.

What you can automate

What these have in common is that they all require reading complex document layouts and pulling exact values out of them. All four are walked through step by step in Use cases.

How is this different from a general AI chatbot?

When one model handles a whole document end to end and the answer is wrong, it is hard to tell where and why it went wrong. Studio separates reading (Parse), sorting (Classify), extracting (Extract), and judging (Instruct) into distinct nodes, so you can check and fix each stage on its own.

Core concepts

• Agent — A pipeline that processes documents through several stages automatically. This is the basic unit you build and run in Studio.
• Node — A functional block responsible for one stage of the pipeline. There are four:
  • Parse — converts any document format into text (details)
  • Classify — sorts documents into your own document types, and splits them if needed (details)
  • Extract — pulls only the fields defined in your schema, as structured data (details)
  • Instruct — takes the results and summarizes, validates, or decides (details)
• Schema — The template that defines which fields to extract from a document. It is the core setting of the Extract node.
• Config — A saved version of an agent's setup. Every change stacks up as a new version, so you can always roll back.

What Studio optimizes for

---

Accounts and pricing

Studio is free to start with. There are only four things to understand: your account, subscription plans, the usage quota, and API keys.

Creating an account

How pricing works

Studio is billed as an organization-level subscription. You pay monthly for the number of seats in a paid organization, and the organization receives a shared usage quota that scales with that seat count.

No features are gated between plans. Parse · Classify · Extract · Instruct, agent runs, schedules, and triggers are available on every plan, Free included. What a plan decides is not which features you can use, but how much you can process.

Page counts are converted at the Document Parse Standard rate ($0.01 per page) for parsing alone. Actual throughput depends on which nodes your agent connects. For a standard workflow of a 10-page document through Parse and Extract, Small covers roughly 52 documents a month and Large roughly 600.

Usage quota

• Shared across your organization. The quota is not split per seat. Your organization's total quota is the per-seat quota multiplied by the number of billed seats.
• Check what is left on screen. The subscription page in your organization settings shows usage and remaining quota for the current billing period.
• Refill cadence follows the plan. Free refills daily; paid plans refill each billing period. Unused quota does not roll over.

What counts against the quota

The same Studio job draws from the same quota whether you run it from the screen or through the API. If you only need document conversion without Studio features, calling the Console API directly is simpler.

Node processing rates follow the published Console price list.

Classify and Instruct do not draw down the quota for node processing during Beta, and we will announce when they move to standard billing. Model usage by Grace conversations and Instruct, however, draws down your organization's quota from September 21, 2026 (UTC), regardless of Beta status. How much is drawn down depends on the model the run uses and the number of tokens it consumes, and some models are not drawn down at all.

When the quota runs out

• Paid plans stop at the quota by default. Credit overage is off until you turn it on, so runs stop once the quota is spent. Turn it on under Settings → Usage in Studio to keep running from your organization credit, and turn on Auto reload as well to buy credits automatically with your registered payment method when your balance drops below your threshold. Only an Organization Owner can change these settings. If overage becomes a pattern, moving up a plan is cheaper.
• Free plans wait for the next refill. Overage is not supported, so runs stop once the quota is spent. The quota refills the next day.

Managing seats

The settled value reflects whichever is smaller, the time left in the period or the quota left in it. Upgrading 3 Small seats to 3 Large seats on day 20 with half the quota spent is charged USD 600 − USD 20 = USD 580. Used quota resets to zero, and the new configuration's quota is granted for the new period. Quota left over from the old period does not carry over. You can check the exact Due now amount in Studio before you confirm the change.

There is no cap on member invitations. Creating an organization, inviting members, managing roles, and consolidated billing are all available on the Free plan. Cancelling a subscription keeps your organization and memberships intact; only the quota returns to Free.

If your member count would exceed your billed seats, a seat is added at the point of invitation and the higher seat count is billed from the next renewal.

Storage

• No capacity limit. Every plan stores uploaded documents and run results without a volume cap.
• Per-file limits still apply. Each file can be up to 500 MB and 1,000 pages.
• Unused data is cleaned up. Data that goes unused for an extended period is deleted. For paid plans the clock starts when a subscription is not renewed or is cancelled. During the retention window after a subscription ends you can still view and download existing data, but not upload new files.

API keys

Running an agent and calling the API both use an API key. A default key is created automatically when you sign up for Studio, so you can start right away without issuing one in the Console.

• Default usage — The key created at sign-up is already connected. No setup needed.
• If you see a "no API key" notice — Use the Create API key button in that dialog to create one without leaving Studio.
• Issuing, naming, or deleting more keys — console.upstage.ai → API keys (keys start with up_).

The same key draws from your Studio quota when it runs a Studio agent, and from Console credit when it calls the Console API directly.

---

Supported file formats and limits

Scanned documents and photos are handled with OCR. If recognition quality is poor, follow the quality checklist in Parse.

---

Processing time

Processing typically takes minutes — not seconds, and not hours. Exact latency depends on document length and pipeline complexity.

---

Where to go next

• Create an agent — three ways to start, plus a tour of the agent builder.
• Nodes — configure Parse, Classify, Extract, and Instruct.
• Use cases — four real workflows you can copy.
• Run and review — run agents, read results, and build ground truth.
• Versions and deployment — ship to production with the Agents API.
• Troubleshooting & FAQ — fixes for the most common issues.