- BATCHED
- ACTIVE
- PENDING
- COMPLETE
Short answer
How can I batch process annotation tasks with my AI?
You can group individual tasks into a single batch and trigger the processing for the whole set with one command. This allows your agent to manage large annotation workloads without manual intervention for every item. You'll see how the batching logic works below.
Batching outcomes
Where your data lands.
What happens when you run a batch command.
BATCHED
Organized task groups
The AI uses the batching tool to pull separate tasks into a single container. You get a structured set of work ready for processing.
ACTIVE
Triggered processing
Once the batch is ready, the AI sends the command to start the work. The tasks move from pending to active status immediately.
PENDING
Awaiting human review
Some tasks in a batch might require a person to step in. The AI flags these specific items for your manual attention.
COMPLETE
Finalized datasets
After the work finishes, the AI closes the batch. You receive a clean, completed set of annotated data.
The workflow
What your AI does when you request a batch.
The AI handles the heavy lifting of organization and execution.
Group tasks
Your agent identifies the specific annotation tasks you want to work on and bundles them together.
create_batchStart processing
The AI sends the instruction to Scale AI to begin working through the newly created group.
create_batchMonitor progress
The agent tracks the status of the tasks within the batch as they move through the pipeline.
finalize_batchClose the set
Once all tasks are handled, the AI wraps up the operation to ensure the data is ready for use.
finalize_batch
Try it
Copy these to start.
Use these prompts to begin batching tasks with your agent.
Starting points
These are starting points. Swap out the specific task names or IDs with your actual Scale AI data.
Scale AI Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_lsC40U9ook5OZ8hrmwvYLqntMTpGHdSfQWvhNSRl/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Scale AI capabilities are ready to use.
{
"mcpServers": {
"scale-ai-mcp": {
"url": "https://edge.vinkius.com/vk_preview_lsC40U9ook5OZ8hrmwvYLqntMTpGHdSfQWvhNSRl/mcp"
}
}
}Create a new batch of annotation tasks from my current project list.
Group these five specific task IDs into a single batch for processing.
Start the processing for the batch I just created.
Finalize the current batch once all tasks are marked as complete.
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Scale AI once, then ask.
Just link your account once. Your credentials stay encrypted, and you can immediately start managing batches through your chat interface.
Connect Scale AI to your AIFAQ
How batching behaves
- 01
Can the AI delete existing annotation tasks?
No. The AI can only create new batches or finalize existing ones.
- 02
Does the AI decide which human annotators get the tasks?
No. The AI manages the batching and triggering, but the actual human assignment is handled by Scale AI.
- 03
Can I add more tasks to a batch after it has been finalized?
No. Once a batch is finalized, it is closed. You would need to create a new batch for additional tasks.
- 04
Can the AI process tasks one by one instead of in batches?
Yes, but using the batching tools is how you group them for efficient, single-command processing.
- 05
Will the AI change my original data files?
No. The AI only interacts with the annotation tasks through the Scale AI interface.
More questions about Scale AI? The Connector page answers them. See everything the Scale AI Connector can do
Connect Scale AI to Claude, Cursor, ChatGPT & more
