Aperture Launches Open-Source AI Code Editor That Verifies Changes Before Application

Aperture has launched as an open-source AI code editor designed to verify edits before they are applied to user files. The tool addresses the common issue where AI-generated code appears correct but fails to parse, resolve imports, or pass tests. Its 'Composer' feature plans changes and stages them as diffs, running five specific checks before application: parsing, import resolution, type checking, preview rendering, and test execution. If a check cannot run, the system reports why rather than marking it as a pass, and it distinguishes between pre-existing test failures and new regressions.

Technically, tests run in a sandboxed Worker within the browser tab, ensuring no network access and zero cost, with execution taking approximately one second. The editor supports node:test, Vitest, and Jest, and attempts one automatic fix if a new failure occurs. Additional features include a design mode for editing CSS or theme tokens via the preview. Aperture is MIT-licensed, free to use, and supports hosted Grok or custom API keys for Grok, OpenAI, Anthropic, Gemini, and DeepSeek. It can also be run locally without an API key.