Launch HN: mrge.io (YC X25) – Cursor for code review

Hacker News Discussion ↗

Perceived value & use cases

  • Many commenters like the direction: AI-focused code review to reduce rubber‑stamping and catch subtle bugs, especially as AI-generated code increases.
  • Solo developers and open source maintainers find value in having a “second pair of eyes” that can be strict without social friction.
  • Some teams report moving from other AI review tools to mrge and seeing better, more useful comments and encouragement of stacked PR workflows.

Comparison with existing tools

  • Mentioned alternatives include Graphite, CodeRabbit, Copilot for PRs, Aviator, and in-IDE agents (e.g., Claude via MCP).
  • Some feel existing AI reviewers are “too nice” and mostly wrong or trivial; mrge’s founders claim better context-awareness and less noise, but at least one user reports poor results on a sample PR (1/11 useful comments).
  • A few users don’t see the need for a separate tool when their AI editor already helps with review.

AI behavior & feature ideas

  • Features called out positively: PR summaries, conceptual grouping of diffs, diagram generation, custom rules inferred from comment history, and conservative one-click fixes.
  • Users want:
    • Rules learned from past reviewer discussions and coding standards docs.
    • Multiple models/personas (security, architecture) and model “promotion” based on accepted feedback.
    • Detection or highlighting of AI-generated PRs.
    • Awareness of previous commits and better handling of large monorepos.

Integrations & workflow

  • Current focus is GitHub; GitLab, Bitbucket, GitHub Enterprise, and local/IDE pre-PR review are frequently requested and said to be on the roadmap.
  • Some worry about having to leave GitHub; mrge clarifies all comments sync back and the web UI is optional.

Security, privacy, and compliance

  • Strong concern over required write/merge permissions; multiple commenters want a read-only mode or branch exclusions.
  • SOC 2 status is important for adoption; mrge states their own certification is in progress and subprocessors are already certified.
  • Some security-minded users still recommend against use until permissions and deployment models (self-hosted/hybrid) are more constrained.

Maturity, pricing, and polish

  • Service is currently free with plans for per-author pricing; trial length and “free” messaging are seen as unclear.
  • Website UX issues (slow fade-in, black screen) and marketing language (“AI era”) draw minor criticism.
  • Overall sentiment: promising and thoughtful, but with open questions about reliability, security model, and ecosystem coverage.