Praxiom AI GitHub integration

Praxiom AI connects to GitHub with OAuth to do two things: create GitHub issues from your recommendations and PRDs, and read the repositories you select so its answers and specs reflect how your product is actually built. It never commits code, opens pull requests or merges anything, and agent-proposed issues wait for your approval.

Updated

What does Praxiom AI read from GitHub?

Praxiom AI lists the repositories your GitHub account can access, but it only reads the ones you select for a workspace (up to 50). When you select a repository, it indexes the code for meaning-based search.

  • Files and folder structure of selected repositories.
  • Code search, both GitHub's keyword search and a semantic index of up to 2,000 files per repository per indexing run.
  • Recent commits and open pull requests, to see what is changing.
  • README and package manifests, to understand the stack and dependencies.
  • Status of issues Praxiom AI created, including a linked pull request if there is one.

Indexing a repository does not use credits. Asking the agent to explore your codebase in depth uses credits like any other agent run.

What does Praxiom AI create in GitHub?

What Praxiom AI writes to GitHub and how each write starts
WhatWhere it startsWhich repository
Issue from a recommendationRecommendations view, create a GitHub issueThe selected repository you pick in the dialog
Issues from a PRDPM IDE, Execute panel, push to GitHubThe first repository selected for the workspace
IssueAsking the agent in chatShown in a preview you approve first

Issues created from PRD tickets carry the ticket's context, acceptance criteria as a checklist, effort estimate, priority and labels. Issues are the only thing Praxiom AI writes to GitHub.

What does it ask before it changes anything?

When the agent wants to open an issue, it records a pending action with a preview of the title and body instead of calling GitHub. The issue is created only after you approve it in Pending approvals or inline in the chat, and you can edit or reject it first. Reading code never needs approval because it changes nothing.

Issues you create yourself from the Recommendations dialog or the Execute panel are created when you click, since the click is the approval.

What does it never do?

  • It never commits, pushes, creates branches, opens pull requests or merges.
  • It never edits, closes or deletes existing issues.
  • It never changes repository settings, collaborators or webhooks.
  • It never reads repositories you have not selected for the workspace.

GitHub's repo permission is broad by design: it is the OAuth scope that allows reading private code and creating issues. Praxiom AI's code only calls GitHub's read endpoints and its create-issue endpoint.

How do you connect GitHub?

  1. Sign in at app.praxiomai.xyz and open the workspace you want to connect.
  2. Go to Settings, then Integrations under Workspace.
  3. Click Connect GitHub and authorize Praxiom AI on GitHub.
  4. Select the repositories this workspace should use. Indexing starts in the background for each new repository.
  5. Create issues from Recommendations or the Execute panel, or ask the agent questions about your code in chat.

Which plans include the GitHub integration?

Every plan includes it: Pro ($19.99 a month), Growth ($89 a month), Power ($149 a month). Sign up or see pricing.

Frequently asked questions

Can Praxiom AI change my code?

No. Praxiom AI reads code and creates issues. It has no code path that commits, pushes, opens pull requests or merges.

Why does Praxiom AI ask for the GitHub repo scope?

GitHub's OAuth repo scope is what lets an app read private repositories and create issues in them. Praxiom AI uses it only for reading selected repositories and creating issues.

Can Praxiom AI create GitHub issues from a PRD?

Yes. Open the PRD in the PM IDE, click Execute, then Generate Tickets, choose GitHub and click Push. Issues go to the first repository selected for the workspace.

Does indexing my repository use credits?

No. Indexing is free. Agent runs that use the code, such as asking how a feature works today, use credits like any other run.

Why should a product manager connect GitHub?

So recommendations and specs account for how the product is built today. The agent can check existing code before it proposes a change, which makes specs for engineers and AI coding agents more precise.

Run your product discovery on autopilot.

Bring your interviews, tickets and surveys. Get scored insights, a PRD and tickets.