Veridu reviews pull requests with two independent AI reviewers from different vendors, currently OpenAI and Anthropic, then reconciles what they found.

A merge is only blocked when both reviewers independently find the same defect and both rate it at or above your threshold. Anything one reviewer raises alone is marked "needs review" for a person to decide. You still see everything; one model's confident guess just can't stop a merge on its own.

Every finding separates what the reviewer checked in the code from what it concluded, so you can verify a claim instead of taking it on trust.

Works on any public PR from the web, or on private repos via the GitHub App, a fine-grained token, the CLI, or as an MCP tool in your editor. Model costs are included, with no OpenAI or Anthropic account needed. The free tier is 5 reviews, no card.

Key features

• Two reviewers from different vendors on each PR, run independently

• Blocks only on defects both reviewers confirm at your threshold

• Single-reviewer findings surfaced as "needs review", never hidden

• Evidence and conclusion kept separate on every finding

• Pick the models on your panel

• GitHub App, web, CLI and MCP; free tier of 5 reviews

Link: https://veridu.io/demo, which shows a real review rather than the home page.