Annota AI
annota.ai
- Free plan
- Web
- Self-hosted
What Annota AI is
Annota AI combines image annotation, document OCR and review, and AI tools in one workspace available on the web or self-hosted. For image work, Magic Select generates an object mask for review, while Magic Annotate finds selected labels in an image or dataset. Users can draw and edit bounding boxes and polygons and export images and annotations in COCO format. Document datasets accept PDF, PNG, and JPEG files; OCR extracts content page by page for review and correction. Document work can be exported as a Markdown ZIP or an Obsidian-compatible LLM Wiki workspace. Compatible AI agents connect through MCP remote servers using Streamable HTTP and OAuth, applying existing organization permissions and using the app’s credits for billable operations. The maker says its infrastructure and AI models run in EU regions without external AI-provider APIs for AI features. The free plan includes one member, two datasets, 4 GB storage, and €4 in monthly credits. Pro is €21.96 a month billed including 22% VAT; Team is €29.28 a month per member, also including 22% VAT. Enterprise deployment is quoted.
Who it suits
Annota AI suits people preparing image and document data for AI, including teams that need shared datasets or self-hosting. Organizations needing YOLO training should note that it requires separate approval.
Good to know, and what to watch
- Exports image annotations in COCO format.
- OCR supports PDF, PNG, and JPEG datasets.
- Free plan includes €4 in monthly credits.
- Supports self-hosting and web access.
- Connected agents use existing organization permissions.
- Free plan is limited to one member.
- Pro includes 20 GB of storage.
- YOLO training requires separate approval.
- Brain is listed as coming soon.
Plans
Every plan and priceEnterprise
Free
Pro
Team
Free plan
The verdict
Annota AI brings image labeling and document review into a single workspace, with export options for both. Compare the plan limits and confirm whether private-beta training or quoted deployment is relevant to your setup.
What the maker publishes
- Brain availability
- Brain, a private knowledge base for questions across document datasets, is listed as coming soon in the plan comparison. source
- Company
- The maker describes itself as an Italian team building tools to prepare data for AI. source
- Document exports
- Document work can be exported as Markdown ZIP or as an Obsidian-compatible LLM Wiki workspace. source
- Document OCR
- Document datasets accept PDF, PNG, and JPEG files; OCR extracts content page by page for review and correction. source
- Headquarters
- Grosseto, Italy source
- Hosting and AI providers
- The maker says core infrastructure and AI models run in EU regions and AI features do not use external AI providers' APIs. source
- Image annotation
- Magic Select generates an object mask for review, while Magic Annotate finds selected labels in an image or dataset. source
- Integrations
- Compatible AI agents can connect through MCP using remote servers with Streamable HTTP and OAuth. source
- Manual annotation
- Users can draw and edit bounding boxes and polygons, then export images and annotations in COCO format. source
- MCP access
- Connected agents use existing organization permissions, and billable operations draw on the same credits as the app. source
- Model training
- YOLO object detection training is in private beta and requires separate organization approval; a paid plan does not grant access. source
- Product
- Annota AI combines image annotation, document OCR and review, and AI tools in one workspace. source
- Security controls
- Workspace permissions govern access, users can revoke connected app authorizations, and SOC 2, ISO 27001, and ISO 42001 programs are in progress. source
Checked against the maker's pages on 7 Oct 2026.