Label Studio hosting

Data labelling for text, image, audio and video in one tool.

ML platformApache-2.0from $11.99/mo

In short

Label Studio hosting at AIHostBox means a dedicated, single-tenant Label Studio instance on our infrastructure: we install it, pin it to a reviewed release, put it behind TLS on your domain, back it up nightly and keep it patched. Plans start at $11.99 per month and the admin account, the data and any API keys are yours.

All modalitiesMulti-annotatorCustom labelling UIPre-labellingExport formats

Label Studio is an Apache-2.0 data labelling tool maintained by HumanSignal, the company that also sells Label Studio Enterprise. One application covers text classification and named entities, image boxes, polygons and segmentation masks, audio transcription, video tracking and time series. Each project’s interface is defined in an XML-like labelling config, so you can combine controls for unusual tasks instead of waiting for a dedicated tool, and templates cover the common cases.

It connects to the rest of an ML workflow in both directions: tasks sync from S3, Google Cloud Storage or Azure buckets, an ML backend can supply predictions for pre-labelling, and finished annotations export to JSON, CSV, COCO, YOLO, Pascal VOC, CoNLL and more. Several annotators need to reach the same instance with their own logins while the data stays in one place, which is the practical reason to run it on a server rather than on one person’s laptop.

Label Studio hosting plans

PlusRecommended minimum

4 GB RAM · 2 vCPU · NVMe

$11.99/mo

Order

ProTeams & heavier load

8 GB RAM · 3 vCPU · NVMe

$18.99/mo

Order

Every plan is a dedicated single-tenant instance with NVMe storage, nightly off-box backups, TLS and managed patching. Quarterly, semi-annual and annual terms are discounted at checkout. Model tokens are not included — you bring your own key.

How we run Label Studio

We follow the official docker-compose: the heartexlabs/label-studio app running under uWSGI, its bundled nginx in front on port 8080, and PostgreSQL for projects and annotations, with the data directory on persistent storage for uploads and exports.

What we configure for you

DJANGO_DB=default with the POSTGRE_* variables pointing at PostgreSQL instead of SQLite
LABEL_STUDIO_HOST set to your https domain so links and file URLs resolve correctly
LABEL_STUDIO_DISABLE_SIGNUP_WITHOUT_LINK so only invited users can register
First account created with LABEL_STUDIO_USERNAME and LABEL_STUDIO_PASSWORD
LABEL_STUDIO_LOCAL_FILES_SERVING_ENABLED and a document root for files you upload over SFTP
S3, GCS or Azure source and target storage connected on request

Which plan fits

Plus (4 GB) suits a handful of annotators working on text or image projects of tens of thousands of tasks. Pro (8 GB) is the better choice for larger teams, long audio or video tasks, and big exports. Any ML backend for pre-labelling runs as a separate service and needs resources of its own.

The honest limits

Reviewer workflows, task assignment, SSO and per-role permissions are Enterprise-only, so the open-source edition treats every user the same. Synchronous exports of very large projects can time out in the UI; the snapshot API avoids that.

What people run Label Studio for

Build a training set

Annotate with several people and reconcile disagreements.

Evaluate a model

Human review of production outputs.

Pre-labelling with your own model

Connect an ML backend so annotators correct predictions instead of starting from blank tasks.

LLM output grading

Rate or rank chatbot responses side by side to build evaluation and preference datasets.

Licence and version policy

Item	Position
Upstream licence	Apache-2.0
Sold as managed hosting	Yes
Tenancy	Single tenant — one customer per instance
Patching	Critical advisories within 24 hours, high severity within 72

Our full position on licences, including the applications we decline to sell, is on the licensing page.

Works well with Label Studio

More ML platform we host

Questions

Label Studio hosting: your questions

What is in the open-source edition and what is Enterprise-only?

The open-source Community edition includes every labelling interface, multiple users, ML-backend predictions, cloud storage sync and all standard export formats. HumanSignal reserves SAML SSO and LDAP, reviewer workflows, task assignment to specific annotators, role-based permissions, activity logs and custom agreement metrics for Label Studio Enterprise.

Which export formats are available?

JSON, JSON_MIN, CSV and TSV for any project, plus COCO, YOLO and Pascal VOC XML for object detection, CoNLL2003 for named entities, ASR_MANIFEST for speech, and brush masks as NumPy arrays or PNG. Very large projects should use the snapshot export API, because a synchronous export from the UI can time out.

Can it read data straight from our S3 bucket?

Yes. Source and target cloud storage for Amazon S3 and S3-compatible stores, Google Cloud Storage and Azure Blob Storage are part of the open-source edition. Label Studio keeps references to the objects and loads them from the bucket when an annotator opens a task, so large image or video sets do not need copying onto the instance.

How do I stop anyone from signing up?

We set LABEL_STUDIO_DISABLE_SIGNUP_WITHOUT_LINK so new accounts need an invitation link from an existing user, and create the first account with LABEL_STUDIO_USERNAME and LABEL_STUDIO_PASSWORD. Everyone who signs up through an invite has the same permissions in the open-source edition.

Where are annotations stored and backed up?

Projects, tasks, annotations and users are kept in PostgreSQL, and uploaded files sit in the Label Studio data directory. Both are included in our nightly off-box backup. If your source data lives in cloud storage, the bucket remains the master copy and only the references are held on the instance.

Run Label Studio without running a server

A dedicated Label Studio instance, deployed and maintained by us. Month to month, cancel whenever.

Browse the catalogue Talk to a human