Skip to content

remote deploy

sinrimin edited this page Sep 18, 2026 · 1 revision

Remote deploy & inventory import

Phase 8 C3 + C4 · designs: inventory & import · job pipeline

Deploy jobs

POST /api/machines/:id/jobs queues a job; the daemon acknowledges and executes it, streaming throttled progress; the machine page shows live state. Jobs survive daemon disconnects with bounded retries (queue → dispatch → ack-timeout sweep → replay). A profile deploy reuses the same plan/apply pipeline as hnx install, fetched by the daemon with a machine PAT — and registers a deployed agent instance on the machine.

Inventory: scan → diff → import

  1. Scan — the daemon reads each agent's native config dirs (claude-code, codex, deepseek, opencode, pi, hermes) and reports a normalized snapshot: MCP rows, skills, commands, agents/hooks where the target has them. Secrets are redacted daemon-side before upload.
  2. Diff — the server compares the snapshot against a chosen profile and shows what would be created or reused.
  3. Import — one click folds the snapshot into platform resources (reuse-or-create) and optionally a new profile.

Capture-as-profile does the same in one step: everything importable from the latest snapshot becomes a zero-or-more-entry profile — an agent's default state is captureable too.

Clone this wiki locally