Armada Multi-Account Fleet Onboarding (Onboarding)

Armada Multi-Account Fleet Onboarding (Onboarding)

1. 01

   Install the signed, notarized Armada app on the client's Mac and confirm the menu bar icon is live

   One $14.99 licence covers every Mac the client owns or controls, so a delivery team can deploy across the whole bench without buying seats per machine. The licence key arrives immediately and carries a 30-day refund window, which gives you room to run a pilot before the client commits.

2. 02

   Point Armada at every config folder the client actually uses

   Verify it is watching ~/.claude plus any ~/.claude-* siblings and CLAUDE_CONFIG_DIR, the Codex home at ~/.codex and CODEX_HOME, and the Grok Build home at ~/.grok and GROK_HOME. Each config folder is treated as a separate account, so a client running three Claude Code logins sees three distinct entries in the unified dashboard.

3. 03

   Confirm the client's own claude, codex, and grok CLI binaries are invokable from the shell

   Armada calls those binaries to fetch plan limits and usage figures. If a binary is missing or shadowed by a version manager, plan limit panels stay empty and the onboarding looks broken to the client.

4. 04

   Record the baseline session inventory before any workflow changes

   Capture which sessions are waiting for input, which are running tools, which are idle, and the context window occupancy per session. This snapshot becomes the before-state you cite in the retainer review.

5. 05

   Stand up the read-only MCP endpoint and its seven tools for the supervisor agent

   The endpoint is read-only by design, so a supervisor agent can query fleet status without any write path into client repos. Document the seven tool names in the client runbook so their own engineers can call them.

6. 06

   Build the alert threshold profile against the client's real plan limits

   Set warnings on 5-hour and 7-day limit consumption with reset times and forecasts visible in the menu bar. Tune thresholds to the client's context budgets so alerts fire before a session stalls, not after.

7. 07

   Walk the delivery lead through fork and account-switch recovery

   Show how to fork a session or continue it on another account when one hits its limit. This is the step that converts Armada from a dashboard into a delivery continuity tool, and it is the demo that justifies the add-on fee.

8. 08

   Hand off the runbook and schedule the first weekly health check

   Leave the client with the config folder map, threshold profile, and MCP tool list. Book the recurring review so the monitoring work sits inside a retainer rather than a one-off install.