Vergo

Execution-backed PR verification.

Vergo is a side-agent CI product: it does not just review code, it reproduces the PR environment, executes verification plans, and posts evidence-rich reports back to GitHub.

Why this exists

AI code review tools are great at static review. They are weaker at proving a PR works in realistic runtime conditions.

Vergo fills that gap by:

• generating a reproducible verification plan per PR
• executing checks in isolated sandboxes
• publishing pass/fail + artifacts + confidence score

Product shape

1. GitHub App receives PR event or /verify command.
2. Planner builds runbook from PR diff + repo context + env requirements.
3. Runner executes plan in sandbox (tests, API checks, smoke checks).
4. Reporter posts evidence comment + check-run status.

Current repo contents (MVP landing)

• src/app/page.tsx — marketing landing page
• src/app/sign-in/page.tsx — sign-in placeholder
• src/app/app/page.tsx — dashboard mock
• .vergo.example.yml — sample per-repo verification config

Startup naming shortlist considered

• Vergo (chosen) — verify + go
• Rivet
• Proofly
• Shipfox
• Runmark

Local development

npm install
npm run dev

Open http://localhost:3000

Next build steps

• Implement GitHub App auth and install flow
• Wire /verify, /verify quick, /verify plan command handling
• Connect planner (pr-repro-skill) as backend service
• Add sandbox runner and artifact storage
• Post PR comments + check-run statuses from real runs