locAGI vs Supervertaler Workbench

locAGI is a managed cloud computer-assisted translation platform combining translation memory, terminology management, AI-assisted translation, and quality review. It is designed for enterprise localization and linguistic teams, including staff and contracted freelancers. The platform is currently in private beta, with public release scheduled for Q2 2027. It lists AI models from OpenAI, Anthropic, Google, Mistral, DeepSeek, xAI, and DeepL, plus private models through Ollama or OpenAI-compatible endpoints. Quality review can use MQM Core, LISA QA Model 3.1, or custom free-text categories, with findings exportable to Excel. Workflows include memoQ MQXLIFF, Trados SDLXLIFF, Phrase MXLIFF, and XLIFF 1.2 and 2.0; more than 40 source formats are supported. A REST API, tokens, webhooks, and an OpenAPI specification support project, segment, QA, and export automation. The platform plan is free, with unlimited seats and included storage; AI is billed separately or can use BYOK. locAGI cannot be self-hosted, and desktop CAT tools are described as more battle-tested for XLIFF files with 50,000 or more segments.

Supervertaler Workbench is a free standalone CAT tool that combines AI translation with traditional computer-assisted translation workflows. It supports AI translation, batch translation, chat, prompts, image context and AI proofreading, as well as local models through Ollama. Listed providers include OpenAI, Anthropic Claude, Google Gemini, Mistral AI, DeepSeek, OpenRouter, Ollama and custom OpenAI-compatible endpoints. Translation features include fuzzy matching, TMX import and concordance search; terminology tools include termbases, automatic term highlighting and TermLens. DOCX and TXT are supported directly, with additional formats through Okapi, including IDML, HTML, XLIFF, PO, XLSX and PPTX. Documentation describes exchange workflows for memoQ, Trados Studio, Phrase and CafeTran. The privacy policy says documents, termbases and translation memories stay on the user's machine, while AI requests go directly to the selected provider when an AI action is triggered. Ollama processes data locally. Workbench runs on Linux, macOS and Windows and supports offline desktop use. Its maker says it is no longer actively developed, though genuine bugs may still be fixed and releases still go out.