diffy.codes

What It Is

Diffy is an AI-powered diff review tool for GitHub and GitLab. It is not positioned as a general-purpose chat-based coding assistant, but rather as a code-change review tool tightly focused on PRs/MRs, commits, and *.diff files. It can generate actionable Markdown feedback alongside a PR and stream the output progressively, making it useful for developers who want to quickly identify risks, improvement opportunities, and potentially missing tests during review.

Core Capabilities

Diffy’s core capability is diff-aware review: it focuses on “what changed in this update” rather than performing a broad analysis of the entire repository. The copy explicitly says it can flag risks, suggest improvements, and call out missing tests or edge cases. On the model side, it uses a Bring your own API key model and supports OpenAI, Azure OpenAI, OpenRouter, Anthropic, Gemini, AWS Bedrock, Cohere, Together, Perplexity, as well as local OpenAI-compatible endpoints, offering a high degree of flexibility. It also supports global instructions and project-level overrides, with a merge toggle for combining contextual rules—useful for teams working across multiple projects.

Pricing and Privacy

The collected text does not disclose whether Diffy itself is paid, whether there is a free plan, or whether a trial is available. The known model is BYOK, meaning users need to configure their own AI provider API key and pay the corresponding model usage costs. On privacy, Diffy says API keys are stored in browser storage and requests are only sent to the provider configured by the user. However, the copy does not clarify whether code diffs pass through Diffy’s own backend, whether logs are retained, or whether enterprise compliance and audit features are supported.

Pros and Cons

Its strengths are a clear integration scenario, covering GitHub/GitLab PRs, MRs, commits, and diff files; streaming output and Markdown rendering also improve the review experience. Support for multiple models and local compatible endpoints helps reduce vendor lock-in. The limitations are mainly around limited disclosure: there is no clear pricing, service support information, Chinese-language capability, or enterprise security details. Review quality also depends heavily on the selected model, prompt rules, and diff context, so AI suggestions still require human judgment.

Best Fit and Access from China

Diffy is best suited for developers, open-source maintainers, and small engineering teams that already use GitHub/GitLab workflows and want to add an AI first-pass review layer to PRs. For non-technical users, configuring an API key and choosing a provider may be somewhat challenging. Access from mainland China is not covered in the copy and will also depend on the availability of GitHub/GitLab and the selected model service. Payment methods are not disclosed either. Alternatives to consider include CodeRabbit, GitHub Copilot code review, Sourcegraph Cody, or manually reviewing diffs with Cursor/Claude Code.