MTurk closes September 30
Moving from MTurk?
Bring your next task to Toloka Platform
Run labeling, evaluation, and data tasks with human contributors. Describe what you need; the AI assistant helps you set it up.
Trusted by Leading AI Teams
What can you run?
Toloka Platform turns a task description into a running project: an AI assistant builds the pipeline with you, vetted contributors do the work, and automated QA checks every result before it reaches you.
Evaluate AI outputs
Get human judgments on model responses and relevance.
Relevance and ranking judgements
Pairwise comparison of outputs
Adversarial red teaming
Collect and label data
Build datasets for training and testing.
Bounding boxes, segmentation, keypoints
Entity, intent and sentiment tagging
Utterance recording and transcript checks
Egocentric video
Review content and generated data
Check content against your quality criteria and policies.
Policy enforcement and abuse flagging
Taxonomy tagging for operations
Authoring and rewriting copy
Organize and enrich records
Make business data more structured and usable.
Field transcription from documents
Open-web attribute sourcing
Entity matching and de-duplication
Need custom data collection or a different workflow?
How it works
What can I bring from MTurk?
Your materials help create a new project. MTurk projects and worker panels do not transfer automatically.
From MTurk
On Toloka
HIT instructions and examples
Attach them as reference material for the assistant to build from.
Tasks and input data
Organize them into a new Toloka project and workflow.
Read the docs
Qualification criteria
Use them to explain the expertise, languages, and other contributor requirements your task needs.
Quality standards
Use your acceptance criteria and examples to define checks for the new project.
Read the docs
Worker panel
Does not transfer. Work is assigned to contributors available through Toloka.
API integration
Adapt your integration to Toloka's API and workflow model.
Build with agents
Bring Toloka into your coding workflow
Fine-tune, compress prompts, collect or label data from the tools your team already uses. Paste the line below into your coding agent.
Results on your own workload within days.
Prefer to build it yourself? Explore the developer docs
Create projects, upload data, run workflows, and export results through the API, TypeScript SDK, and CLI. Adapt your existing MTurk integration — the Pipeline API is in beta, and request and response formats may change.
Quality checks,
built into the work
Quality control is configured on the pool — the group of tasks that shares the same settings.
89.1% accuracy catching failures — before they reach your pipeline.
Overlap
How many contributors independently complete each item.
Contributor tiers
Global crowd, general annotators, or domain experts — matched to the work.
Review layers
Automated review on every output, with human review where the work needs it.
Work runs on a network of 200,000+ contributors across 90+ domains, and rejected expert work returns for correction or reassignment.
Explore quality controls
The assistant drafts quality criteria from your task description. Review what counts as a pass or fail, add examples of difficult cases, and adjust the checks before launching.
Where your task needs comparison or agreement checks, collect independent answers from several contributors on the same item. An LLM reviewer checks submissions against your criteria, and supported expert-labeling workflows can add human review on a configurable share of submissions.
Shopify's segmentation skill-judge accuracy improved in one month. Toloka's human evaluation supported this customer-specific result through prompt engineering.
Start small.
Know what to expect.
Start within hours
Projects typically begin within hours. Simple classification can take 24–48 hours; complex work may take longer.
No minimum spend
Review your batch estimate, including labeling and quality checks, before launching.
50 minutes
Lovi.Care's setup-to-launch time. Test scenarios were completed and labeled in under a day. A customer example; timing varies by project.
Spend $50.
Get $50 in Toloka credit.
Credit is added after your first $50 of qualifying spend clears.
Moving a complex workflow?
High volumes, custom interfaces, specific contributor requirements, ongoing automation, or compliance and data-handling needs — these are worth talking through before you rebuild them.
In a 15-minute call we'll review what you run today, identify what has to change, and tell you which parts fit self-service and where you need support.
Questions before you move
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><path d="M 12.5 15.5 L 0 15.5 L 0 12.5 L 12.5 12.5 L 12.5 0 L 15.5 0 L 15.5 12.5 L 28 12.5 L 28 15.5 L 15.5 15.5 L 15.5 28 L 12.5 28 Z" fill="rgb(0, 0, 0)" height="28px" id="v_HNFUCmW" transform="translate(10 10)" width="28px"/></svg>)
What should I save before MTurk closes?
Your HIT instructions and task interfaces, qualification criteria, examples of accepted and rejected work, input data and results, and your transaction and account records. Check Amazon's notice for the deadlines on outstanding assignments, approvals, bonuses, and refunds.
Can I import projects, qualifications, or my worker panel?
No. Attach your existing instructions, criteria, and examples as reference material, and the assistant uses them to build a new Toloka project for you to review. Worker panels and qualifications do not transfer.
Can Toloka support surveys or specific participant requirements?
Fit depends on your study, participant requirements, and research constraints. Tell us about the sample, screening criteria, and work you need before choosing a setup.
How much does Toloka Platform cost?
Project price is calculated after the setup is clarified, based on what the project needs. No hidden fees, no minimums, and no contracts.
Do I need technical experience?
No coding or data science expertise required. Describe your task in plain language and the AI assistant asks clarifying questions, configures quality controls, and recommends the right contributor tier. There is no onboarding — sign up and start.
What makes the contributor tiers different?
Domain experts are credentialed specialists in fields like medicine, law, and finance. AI tutors are RLHF-trained annotators who follow complex rubrics for preference labeling and instruction tuning. General experts handle high-volume work such as classification and categorization. The AI assistant recommends the mix for your task and budget.
How do I pay for projects?
Add credit by card, processed by Stripe. If your company requires bank transfer, contact our team.
How does the credit offer work?
For eligible new accounts, spend your first $50 on tasks and receive $50 in Toloka credit after that spend clears. One match per account. See the offer terms for eligibility and expiry.
Where can I get help with my migration?
Talk to a migration specialist about volume, deadlines, integrations, and other requirements. We'll help assess the self-service fit and identify where you need support.
Ready for your first batch?
Bring your instructions. Review the setup. Launch when you're ready.