Veriva-AI installs statistics

The verification layer for AI-generated code. Independent checks on every PR.

The verification layer for AI-generated code

As more of your codebase is written by Claude Code, Cursor, Copilot, and the next

five agents, who actually verified that the code does what was asked becomes an

architectural question — not a vibe.

Veriva runs an independent check on every pull request. Ten governance stages,

multi-model routing across Claude Haiku / Sonnet / Opus, per-agent identity

tracking, deny rules that override AI judgment, and a full provenance trail on

every finding.

We don't generate code. We verify it.

What you get on every PR

• A trust score (A–F) that reflects the actual verification outcome — not how

confident the model sounds.

• Per-agent attribution. Claude Code, Cursor, Copilot, or human — line by

line, with a 90-day rollup per author.

• Slopsquatting + supply-chain detection. We flag hallucinated packages

before they ship.

• Custom policy gates. Declare your rules once in .veriva/rules/*.yml;

they fire on every PR with a full audit trail.

• A merge gate that's auditable. Five tiers from Advisory to Locked, with

override reasons captured per merge.

Pricing

• Hobby — Free. 3 private + unlimited public repos, 3 members,

250 KB reviewed/month on private. Install directly from Marketplace.

• Team — $99 / month flat + 2,000 KB included ($0.05/KB overage).

Unlimited repos + members. Upgrade at

veriva.dev/pricing.

• Scale — $499 / month flat + 20,000 KB included ($0.05/KB overage).

Cross-PR dedup, unlimited auto-fix, 90-day audit log.

• Enterprise — Custom. SSO/SAML, self-hosted, signed DPA + MSA.

Contact sales.

Description from the GitHub Marketplace listing.