Skip to content

GitHub agent-sync

Keprix mounts the durable-memory bridge across Hermes, Verlox, and agent environments (canonical repo: malike2356/agent-sync).

Configure in the GUI: Settings -> Agent sync (/settings/integrations/agent-sync), or the Settings overview card.

Canonical repo: malike2356/agent-sync (changeable in the GUI).

What it is

  • Shared markdown/skills memory over GitHub
  • Pull on an interval; push on approved durable writes (or interval)
  • Local keyword index under ~/.keprix/data/github-agent-sync/<scope>/
  • Policy denies secrets and ephemeral chat dumps

Not the Obsidian vault. Vault sync uses Syncthing. Do not point agent-sync and Syncthing at the same write-heavy tree.

GUI setup

  1. Open Settings -> Agent sync
  2. Paste a fine-grained GitHub PAT
  3. Set owner/repo/branch (defaults are fine)
  4. Set product to keprix (use hermes on Fowler)
  5. Enable, Save, then Pull now

No .env changes are required for normal use. Tokens saved from the GUI live under ~/.keprix/data/github-agent-sync/ (mode 600).

API (same as the GUI)

  • GET /api/agent-sync/status
  • PUT /api/agent-sync/settings
  • POST /api/agent-sync/pull|push|index|search|note

Tool

github_agent_sync (toolset github-agent-sync): actions status|pull|push|search|note.

Hermes / Fowler

Local Hermes (Fowler) should use product hermes on the same repo. VPS Keprix uses keprix. They share memory/ and skills/ while keeping distinct agent mounts.

Optional env bootstrap

Env vars exist only as an optional bootstrap for headless hosts. Prefer the Settings GUI whenever you have a browser.