Harper vs Trinka

Harper is a free, open-source grammar checker that flags capitalization, spelling, awkward phrasing, and grammar errors. The maker says checks run locally, with no cloud round-trips, telemetry, or LLM in the loop, and that suggestions appear in under 10 ms. Harper is available through browser and editor extensions, a language server, a JavaScript library, a Rust crate, and native apps. Listed integrations include Chrome, Brave, Edge, Firefox, VS Code, Cursor, Obsidian, WordPress, Neovim, Zed, Helix, Emacs, and Sublime Text. Its language support covers English and British, American, Canadian, Australian, and Indian English dialects. Harper Desktop lets users toggle rules, add dictionary entries, select a dialect, and turn checking on or off per app. The desktop app is labeled beta for macOS 14 or later, and the Windows desktop app is also described as a beta; its binary was not signed at the time of the listed post. Harper is free, with no account or credit card required.

Trinka is an AI grammar correction and language enhancement assistant for academic and technical writing. It checks grammar, suggests paraphrases, reviews consistency and can proofread uploaded MS Word documents for download. Other tools include AI-assisted writing, citation features, plagiarism checks and a journal finder. Trinka Cloud checks English and Spanish grammar and detects the language automatically. Its browser plugin supports writing in Google Docs, Authorea, Evernote and Medium; listed platform options also include web, API, desktop and self-hosted access. Institution plans list LMS and SSO integrations such as Canvas, Moodle, Blackboard and D2L. The free Basic plan does not expire, but monthly usage limits apply. Premium costs $20 a month, billed monthly, and includes up to 120 proofread files per year and 25 AI writing requests per month. Confidential Data costs $300 a year with annual billing. Trinka says user data is encrypted in transit and at rest. Its Trust Center says paid-plan data is not used to train AI models and Confidential Data Plan data is deleted after processing.