What Skene connects to.
GitHub shows what the code records. Supabase grounds the journey in your database. PostHog or Mixpanel supplies analytics counts. Each adds evidence to the plan without replacing the tools you already use.
Each connection is scoped when you make it.
The repositories you select
Choose selected repositories or all repositories when you install the GitHub App.
Access mode set at connect time
Supabase read-only and read-write are separately registered OAuth applications.
Eleven tools over MCP
Hosted by Skene, for coding agents.
Each connection is scoped when you make it.
The repositories you select, the access mode you pick, the API key you issue.
None of them is needed to write your first plan. GitHub is the one you need to implement it, and every further connection makes the answer more accurate.
Suggested fixes stay in a pull request for your team to review. Supabase starts read-only; changing access mode requires a separate permission grant. Analytics connections read your existing tools.
Two providers, read as sources.
Exactly two analytics providers connect. PostHog, over its HogQL query API, and Mixpanel, over its service account query API. Skene reads the counts on demand behind a five minute cache.
What the installation reaches.
The installation token does three things.
- Lists the repositories you picked
- Fetches the skene-context/ files the engine reads
- Posts pull request reviews
GitHub is the only repository host. GitLab and Bitbucket are not supported. Checks respond to push, pull_request and issue_comment.
Production and development keep their own place.
Development and production databases are assigned independently. If you have not assigned an environment, its checks report as not linked instead of passing.
- Deploys, triggers and every other write go to the production row only
Connect one Supabase project.
Read-only and read-write are two separately registered OAuth applications, not a scope on one. Changing it later means reconnecting, not widening a grant you already gave.

