Limina
Context-aware de-identification for PII, PHI, and PCI in structured or unstructured data, with support for 52 languages.
About Limina
Limina helps teams de-identify sensitive data without discarding the context needed for analysis, reporting, or model training. It can identify, redact, or replace PII, PHI, and PCI in structured and unstructured content, covering more than 50 entity types and 52 languages. Its context-aware approach addresses cases that pattern matching can miss, such as names spread across sentences or payment details spoken over several turns. Access is available through an API, with SaaS processing, containers for an on-premises environment or VPC, and marketplace deployment through AWS, Snowflake, and Azure. Common workloads include clinical notes, legal documents, call-center transcripts, migrations, model-training data, and breach reporting. The Free plan allows up to 75 API calls each day, subject to rate and file-size limits. Batch handles large one-time jobs with volume pricing. Enterprise adds regional or hybrid deployment, multi-cluster support, pinned model versions, custom SLAs, and committed-volume pricing. Costs beyond the Free plan require a quote, and specialized data may need integration work and detection tuning before production use.
Best for
Best use cases
A clinical-data pipeline removes identifiers from notes before analytics or model training.
A migration or breach-reporting project processes a large dataset as a one-time batch.
A call-center workflow detects sensitive details that span multiple conversational turns.
Compare before choosing
Choose Limina over pattern-only tools when context across sentences or turns affects detection.
The free tier suits validation, while sustained or high-volume production requires quoted pricing.
Enterprise deployment offers more infrastructure control than SaaS-only processing but adds implementation work.
Features
Sensitive-data analytics
De-identify PII, PHI, or PCI while retaining useful context for downstream analysis and reporting.
Model-training preparation
Prepare restricted structured and unstructured datasets for model training without exposing identified records.
Large batch processing
Process migration, breach-reporting, or dataset-monetization workloads as a pipeline-ready one-time batch.
Pros
- Context-aware detection handles sensitive details that span sentences or conversational turns.
- Support for 52 languages and more than 50 entity types reduces the need for separate tools.
- SaaS, VPC, on-premises, regional, and marketplace options support varied residency requirements.
Cons
- The free tier is capped at 75 API calls per day and imposes file-size limits.
- Batch and Enterprise costs are quote-based, which makes upfront budget comparison harder.
- Production onboarding and accuracy tuning add implementation work for specialized datasets.
Pricing Plans
Free
- Up to 75 API calls per day
- Rate limit of 10 calls per minute
- 250KB for txt, json, csv, and xml; 4MB for other file types
- Public documentation
Batch
- Process a large dataset in one run
- Pipeline-ready output
- Self-serve onboarding
Enterprise
- Committed-volume pricing
- Hybrid or regional deployment
- Multi-cluster and multi-region support
- Pinned model versions
- Dedicated customer success manager
- Guided onboarding
- Flexible payment options
- Custom SLA
* Pricing information is for reference only. Please refer to the latest announcement on Limina's official website.