Introduction

Eblan is powered by Claude, Anthropic's frontier AI model, and works with any language or framework on GitHub. There's no dashboard to check and no CI config to maintain: install the app, open a PR, and Eblan takes it from there.

Quickstart

Getting your first review takes about a minute:

1. Install the GitHub App. During the private beta, installations are invite-only — join the waitlist and we'll send you an invite with the install link.
2. Pick repositories. Choose which repos Eblan should review. You can change this any time from the app settings on GitHub.
3. Open a pull request. Eblan reviews it automatically — no comment, label, or command needed.

That's it. Most first comments land within two minutes.

Configuration

Eblan works with zero configuration. When you want more control, add a .eblan.yml file to your repository root:

# .eblan.yml
review:
  depth: thorough      # "fast" skips deep analysis on trivial diffs
  max_comments: 10      # cap comments per PR to avoid noise

ignore:
  paths:
    - "**/generated/**"
    - "**/*.lock"
    - "docs/**"

conventions:
  read: true            # learn from CONTRIBUTING.md and similar files

fixes:
  suggest: true        # attach one-click suggested fixes to comments

Configuration changes apply to the next pull request — no restart needed.

How it works

When you open a pull request, Eblan runs a short pipeline:

1. Collect. Eblan fetches the diff, the surrounding code, and relevant context (conventions docs, related files) through the GitHub API.
2. Review. The changeset is sent to Claude with carefully structured instructions and your repo's conventions. Claude reasons about what the change does — not just what it looks like.
3. Report. Findings are filtered for confidence, ranked by severity, and posted to the PR as regular review comments with suggested fixes where possible.

Eblan is tuned to prefer precision over recall: it would rather say nothing than nitpick. In beta, about a third of its comments flag an issue a human reviewer agreed was real.

Privacy & security

Your code is yours. Eblan's data handling is deliberately boring:

• Code is fetched for review, sent to the Claude API, and never stored by Eblan beyond lightweight PR metadata.
• Code sent to the Claude API is never used to train models.
• All traffic is encrypted in transit; GitHub permissions are least-privilege (read code, write PR comments).

Read more on the security page.

FAQ

Which languages does Eblan support?

All of them, in principle — Eblan reviews the change, not the syntax tree. It's most battle-tested in TypeScript, Python, Go, and Rust.

Does Eblan approve PRs automatically?

By default Eblan leaves comments and an approval when it finds no blocking issues. You can require a human approval on top — Eblan never merges anything.

Can I trigger a review manually?

Yes — comment @eblan review on any PR to re-run the review, or @eblan explain on any comment to get a deeper explanation.

Something's missing. Where do I ask?

Email hello@eblan.site — during the beta, the founders answer every message.