Skip to content

Recipe Monitor Multiple Machines

nguyen.david edited this page Aug 13, 2026 · 1 revision

Recipe: Monitor Multiple Machines

Use when: one dashboard should cover local workstations, servers, WSL environments, or other SSH-accessible hosts.

Design first

Choose one authoritative CCAM writer and give each producer a stable source identity. Decide per host whether you need near-live hooks, periodic SSH mirroring, or a one-time import.

Procedure

  1. Deploy the central dashboard with TLS and separate dashboard/hook tokens.
  2. For near-live producers, configure CCAM_DASHBOARD_URL and a file-backed hook token, then install hooks on that host.
  3. For SSH collection, establish key-based access from the CCAM service account and pin/verify host keys.
  4. Add one Remote Data Source at a time and run Test.
  5. Sync and confirm Claude and Codex provider statuses independently.
  6. Verify imported sessions carry the intended source and project path.
  7. Create a short test session on that host and measure visibility delay.
  8. Add source-specific monitoring and document ownership.

Avoid duplicates

Be deliberate when a host uses both forwarded hooks and SSH mirroring. Stable session identity normally permits enrichment, but path/home mismatches can make one run look like two. Test with a known session before scaling out.

WSL note

CCAM can mirror WSL-hosted CLI data through the Windows SSH host using the supported remote-source path. Validate path identity and permissions with one provider before adding the second.

Done when

Every host has a named owner, collection method, last successful test, expected delay, and recovery procedure. The central instance still has exactly one SQLite writer.

See Remote Machines and Data Sources.

Clone this wiki locally