Skip to main content
Open-source practice feedback for software teams

Learn from the work you’re already doing

Hephaestus reviews the pull requests, issues, and code reviews your team already writes against the engineering practices you choose. It gives each developer feedback that cites their work and names a next step.

Feedback appears on the work itself, on each developer’s private practice page, and in a conversation. There, developers can ask why, ask for an example, or disagree. Workspace admins can approve feedback before it appears on the work.

Four cards from one change: an issue, a pull request, a review and the merge. Each card has the practice feedback that Hephaestus attached to it.Four cards from one change: an issue, a pull request, a review and the merge. Each card has the practice feedback that Hephaestus attached to it.

Pick the guide that matches your role

Every page says who it is for in its first line.

You use Hephaestus

User guide

What the feedback is, where it reaches you, and how to talk it through.

You run a deployment

Admin guide

Install it, connect a source and an AI provider, and keep it healthy.

You change the code

Contributor guide

Set up the services, follow the standards, and ship a change.

Features​

Practices

Start from a curated catalog or write your own. Each states what good work looks like.

Practice catalog

Feedback on the Work

A comment on the pull request, merge request, or issue.

Practice feedback

Practice Page

Each developer’s private view of feedback and progress per practice. Workspace admins see the observations and delivery status behind it.

Practice profile

Conversations

A weekly digest you can turn on arrives in the conversation. Developers ask why, ask for an example, or disagree.

Ask Hephaestus

Feedback Mode

Off, Approve first, or Automatic, per workspace, practice group, or practice. Approve first holds feedback on the work until a workspace admin approves it.

Practice review

Your Stack

Works with GitHub, GitLab, Slack, and Outline. Your own model provider: OpenAI, Azure OpenAI, or any OpenAI-compatible endpoint. Helm on your Kubernetes.

Install with Helm