Human in the Loop API

Human in the Loop API
for AI Teams

Connect real people to your AI pipeline. Send model outputs, data or questions through the API, and get human reviews, feedback and labels back from workers in the countries you choose.

Human feedback APIAny country and languagePay per task

200+

Countries

13,487

Tasks completed

195,000+

Registered workers

~4 Hours

Avg time to results

How It Fits

Add Real People Where Your Model Needs Them

Keep automation for the easy cases and send the rest to people. Their answers flow back into your system.

01

Your model or app

Your system decides which items need a human check.

02

RapidWorkers API

You create a task with instructions, countries and pay.

03

People review or create

Workers complete the task and submit answers or files.

04

API returns results

Your code fetches submissions and reads each answer.

05

Your system updates

Approve good work and feed results back into your pipeline.

Use Cases

What Teams Send to People Through the API

Low-confidence predictions

Send items your model is unsure about to people for a final decision.

Human feedback for training

Collect comparisons and ratings of model outputs as human feedback for fine-tuning and RLHF.

Ongoing model evaluation

Run human evaluation automatically after every model release and track the results.

Data collection on demand

Request new speech, images or text from real people whenever your dataset needs more examples.

Content and data checks

Ask people to verify records, labels or generated content before it goes live.

Edge cases and failures

Send examples your model gets wrong to people for labeling, so you can retrain on them.

Capabilities

What You Can Do Through the API

Create tasks

Send instructions, countries, pay and the number of submissions.

Attach reference files

Add campaign images or AI-task reference files to a task.

Check status and progress

List campaigns and read remaining slots and pending submissions.

Get submissions

Fetch worker answers, proof text and uploaded files.

Approve or reject

Approve good work, reject the rest, or run bulk actions.

Extend a task

Add more submissions to an AI task without creating a new one.

How It Works

From API Call to Human Answer

Create tasks from your code, let real people answer and feed the results back into your model or product.

Get API access
  1. Create a task from your code

    Send the instructions, the items to check, the countries and the pay per submission.

  2. People complete the task

    Workers in the selected countries read your instructions and submit their answers or files.

  3. Get the results

    Your system fetches the submissions and reads each worker's answer.

  4. Approve and learn

    Approve good submissions and reject the rest. Feed the human answers back into your model or product.

Good to Know

Design Your Human-in-the-Loop Flow

Small batches

Send items in small batches inside each task.

Clear rules

Use clear rules and examples, the same way as in manual tasks.

Several people

Ask several people when the decision is subjective.

Keep a fallback

Keep a fallback in your app while human answers are pending.

Not real time

Use the API for review queues, batch feedback and dataset updates — not for decisions users wait for in real time.

Minutes to hours

Turnaround depends on task length, countries and pay. Plan for async results.

How We Work

Integrate Yourself or Let Us Help

Self-serve

Integrate the API yourself

Use the API documentation to create tasks and fetch results from your own code.

Read the API docs

Fully managed

Let our team set it up

We help design the task flow, set quality rules and run a pilot with you before you scale.

Talk to us
Pricing

Estimate Your API Usage Cost

Tasks created through the API are priced the same way as on the platform: you set the pay per task, plus the 15% platform fee.

?
?
min
?
$
FAQ

Frequently Asked Questions

An API that lets your software send tasks to real people and get their answers back automatically. Teams use it to review uncertain predictions, collect human feedback for fine-tuning and RLHF, and verify data.

Create tasks with instructions, countries and pay, attach reference files, check progress, fetch answers and files, approve or reject in bulk and extend AI tasks. It works with your employer account.

Usually within minutes to hours, depending on task length, countries and pay. The API is built for review queues and batch feedback, not real-time decisions, so keep a fallback in your app.

Yes. Set the countries when you create the task and state the language in your instructions.

You approve or reject each submission, can ask several people per item and can add control items with known answers.

Tasks cost the same as on the platform: you set the pay per task, starting from $0.10, plus a 15% platform fee.

Yes. Our team can help design the task flow, set quality rules and run a pilot with you before you scale.