Skip to content

GitHub

Liddr reads issues, pull requests, and source files from your repos during ingest (Know and Signals), and — for reverse sync — pushes branches and opens pull requests when a Forge build completes. Both directions authenticate with a GitHub personal access token.

What Liddr accesses

DirectionAccess
Ingest (read into your knowledge graph)Read-only
Reverse sync (Liddr writing back)Read & write — only when a Forge action publishes back to this source

Setup

  1. In GitHub: Settings → Developer settings → Personal access tokens → Fine-grained tokens → Generate new token.
  2. Under Resource owner pick your org; under Repository access select the repos Liddr should work with (or All repositories).
  3. Grant the permissions listed below and generate the token.
  4. In Liddr: Settings → Connections → GitHub, paste the token, then Test and run an Ingest. If a sync fails on permissions, the ingest history names the missing scope.

Required scopes

Ingest — minimum

  • Metadata: Read (mandatory for any API access)
  • Contents: Read (source files)
  • Issues: Read
  • Pull requests: Read

Reverse sync — additionally

  • Contents: Read & write (push build branches)
  • Pull requests: Read & write (open / close PRs)

TIP

Only grant write access on the repos Forge builds target. Add Workflows: Read & write only if builds may edit files under .github/workflows — GitHub rejects such pushes without it.

Classic token

Ingest — minimum

  • repo (covers private-repo read; use public_repo if everything is public)
  • read:org (lets org repo listing see team-scoped private repos)

Reverse sync — additionally

  • repo (same scope — classic tokens don't separate read from write)

TIP

repo + read:org is also the maximum Liddr ever needs — never grant admin:*, delete_repo, or workflow unless builds edit CI files.

Reference

Liddr — grounded in your reality, linked to the evidence.