Introduction

Lara is a new adaptive translation AI that combines the fluency, reasoning, context handling and instruct capabilities of LLMs with the low hallucination rate and latency of MT. On top of this, Lara is adaptive which means that Lara does not require training, instead is able to adapt to any domain on the fly leveraging previously translated content or context.

Whether you need standalone integration through our SDK or seamless LLM enhancement via MCP, Lara provides powerful translation capabilities to elevate your multilingual applications.

SDK

Build custom translation AI with complete control.

Our SDK empowers you to build your own branded translation AI leveraging our translation fine-tuned language model. All major translation features are accessible, making it easy to integrate and customize for your needs.

The Lara SDKs allow you to programmatically translate text, optionally providing context and translation examples (translation memories) in order to further improve quality.

Available Programming Languages

The SDK is available for Python, NodeJS, Java , PHP and GO. All SDKs expose the same methods and will rely on the same features.

Ready to build with Lara? Get Started with the SDK and start creating your custom translation experience today.

Why SDKs and not REST APIs?

Although low level REST APIs can be made available for specific use cases (contact us), we highly recommend using the SDK for the following reasons:

• They are better documented.
• The SDK are easier and faster to integrate.
• They implement enterprise grade security natively. Requests are encrypted and signed.
• They will include the future audiovisual features like audio-to-audio, subtitles and dubbing that will not be available in the only-text REST APIs.

Lara Models

The Lara API provides multiple translation models optimized for different trade-offs between quality, latency, and cost.

Lara Base is the default model and offers the best balance between speed, cost, and latency. It is recommended for most general-purpose translation workflows.

Lara Think is an advanced, quality-optimized translation model designed for workflows where linguistic accuracy and consistency are critical.

It performs multi-step linguistic analysis across all available project assets, including glossaries, style guides, and contextual metadata. This deeper analysis enables automatic detection and correction of approximately 80% of major linguistic issues, resulting in improved terminology adherence and stylistic consistency compared to standard models.

Both Lara Base and Lara Think are currently available via the SDK.

Additional Models

In addition to the two models above, the Lara API also offers two additional options for high-volume users:

• Lara Flash which delivers an impressive median latency of 50ms.
• Lara Batch which is the base model optimised for cost, up to 10x less.

If you’d like to access any of these models, please contact our team.

MCP Server

Lara is also available as a MCP-compatible server exposing its translation capabilities via the Model Context Protocol (MCP). In this setup, Lara acts as a specialized translation agent, seamlessly integrating into MCP-aware environments like Claude, ChatGPT, or any other system that supports MCP agents.

By integrating Lara directly into LLM workflows, you gain access to superior translation quality for domain-specific content, the ability to maintain context and terminology consistency across multilingual conversations—all without disrupting the user experience or requiring external API calls. Find out more about why to use Lara inside an LLM or to set it up and start translating, check out the Getting Started guide for the MCP Server

Subscription Options

Both the SDK and MCP server work seamlessly with any Lara Subscription (Free, Pro and Team), whether you're building standalone applications or enhancing AI workflows. For advanced functionality, expert-level APIs are available by requesting a customized enterprise plan.