MacRun Agent Plan Provisioning and Client Codebase Handoff (Onboarding)
A sequence with 8 steps: Choose the MacRun plan type at signup: agent plan for 24/7 Claude Code and Codex work, or CI runner plan for macOS build pipelines.
By InnovaAI ResearchPublished Updated
What are the steps?
MacRun Agent Plan Provisioning and Client Codebase Handoff (Onboarding)
- 01
Choose the MacRun plan type at signup: agent plan for 24/7 Claude Code and Codex work, or CI runner plan for macOS build pipelines
The two plans bill differently. Agent plans run autonomous coding sessions; CI runner plans (M4 16GB/256GB at $119/month, M6 16GB/256GB at $139/month) meter unlimited macOS build minutes with no per-minute charge. Pick one per client engagement so the retainer line item maps to a single invoice.
- 02
Select the hardware tier and confirm the pre-order window before promising a client start date
The M4 16GB/256GB runner is a pre-order with roughly one week lead time. M6 and M5 Pro tiers are the higher-throughput options for heavier agent workloads. Tell the client the real ship date, not the signup date.
- 03
Sign in with your own Claude or ChatGPT account credentials inside the MacRun browser desktop
No API key is required. The machine holds a persistent logged-in session so agents keep running after you close the tab. Store the account credentials in the agency password manager, not in a shared client doc.
- 04
Verify the dedicated Mac mini stays awake and reachable through the browser desktop session
Confirm the persistent logged-in session survived a browser close, and that you can reopen the desktop to approve a login or intervene when an agent stalls. Wired ethernet and battery backup are part of the rack setup; test the session before you hand anything to a client.
- 05
Connect the client codebase and run one scoped coding task through Claude Code or Codex
Point the agent at a single bounded task first, then confirm the output lands as a pull request the client can review. This is the proof step that justifies the retainer before you scale to multi-repo work.
- 06
Attach the repo to GitHub Actions if the client needs macOS builds
MacRun runners are GitHub Actions ready, and the same hardware works with GitLab, Bitbucket, Buildkite, and Jenkins. Pin the Xcode version and warm the local SSD cache so build times stay predictable across the client's release cycle.
- 07
Install any client-specific tooling via Homebrew or sudo on the MacRun machine
Custom dependencies go on the dedicated instance, not a shared runner. Document every install in the client runbook so a second agency operator can reproduce the environment.
- 08
Hand off runner access credentials and the runbook, then log the monthly cost against the client retainer
Give the client read access to build results and PRs, keep admin access inside the agency. Record the $119 or $139 monthly line so the retainer margin is visible at renewal.
More on MacRun
- StrategyMacRun Turns Idle Agent Hours Into Billable Retainer Capacity
- ConceptMacRun Client Fit Matrix
- Evaluation RuleWhen to Adopt MacRun: Persistent macOS Agent Infrastructure, Not Occasional Builds
- Decision FrameworkMacRun: Buy vs Skip (Dedicated Mac mini for AI Coding Agents and macOS CI)
- Failure PatternThe MacRun Idle Rack Trap: Why Agencies Fail With MacRun When Agents Run 24/7
- Implementation BlueprintMacRun iOS CI Runner Migration (5-7 days)