Case study · Both
Keeping AI-assisted development projects ready to resume
How I built Command Centre to keep project history, handoffs and action status connected when resuming AI-assisted development.
In brief
The next step in a project can be buried in yesterday’s session or an old report. Command Centre connects that context in a local working view. Historical reports stay intact while action status changes separately, and I choose which work to start after checking the evidence.
Who this can help
An independent developer, consultant or small-business technical owner coordinating several AI-assisted projects.
The problem
Resuming a project often starts with reconstructing what happened last time. When several AI-assisted sessions are involved, I need the source of a next step as much as the next step itself.
Command Centre is the local working tool I built around that problem. It brings project context, development activity, reports and reusable routines into one place so I can inspect the evidence before starting the next task.
I shaped it around the records and habits I already relied on. A session recap, a project note and an action list each have a different job. Keeping those meanings intact was part of making the tool fit the work.
What I built and how it works
The Activity view reads session recaps and their recorded next steps. Projects combines the project registry with folder and version-control state. Briefings presents reports, while Radar keeps the status of each report action separately.
That separation is deliberate. A report records what was known when it was written. An action can later be completed, dropped or left open without rewriting the historical report. Each action keeps the identity of its source report, so similarly numbered actions don’t become the same task by accident.
The system can also identify leads suggesting an old next step may already be resolved. Those leads give me somewhere to look. Reconciliation still requires checking what the action actually asked for and what the later work completed.
Other views cover routines, AI-client connections and reusable skills. Launching an implementation task opens an agent session with the requested work. The dashboard gives me a way to start that work and return to its context.
Where AI and vibe coding helped
AI-assisted development and vibe coding helped me build a tool for managing the work itself. Command Centre organises session evidence and can launch agent tasks. Its project views and status records do not require a model to invent or infer that work has completed.
What people still control
I choose the action to start and review its result. A launch button doesn’t establish that a change was completed, committed or deployed. Historical activity and time views remain read-only views of their source evidence.
The application runs locally. Some connections and routines require a manual restart or a platform-specific step, which the interface needs to make clear.
What the evidence supports
The supported result is a working view of project continuity: what happened, what was left, which report raised an action and where its status belongs. I can inspect those records before asking an assistant to continue, then give it the relevant context.
The Time view adds another kind of evidence by grouping recorded Claude Code activity around idle gaps. That can help me understand a project’s session history, provided I read the figures in context.
Limits to keep in view
Overlapping sessions can produce overlapping totals. Those figures aren’t exclusive human working hours, billing records or proof of productivity gains. Mac support also has remaining routine coverage work. Command Centre is an internal tool; this case doesn’t claim a commercially packaged team product.
Frequently asked questions
Does launching an agent task mean the work is finished?
No. Launching starts a session. Completion, a commit and a deployment each need their own result evidence.
Why keep action status separate from the report?
The report preserves what was known at that time. A separate status can change without rewriting the historical source.
Can recorded session time be used as billable hours?
The view contains activity spans, including overlapping sessions. It does not establish exclusive human labour or billing accuracy.
A useful next step
Choose a project you regularly struggle to resume. I can help map the handoff, source records and decisions that would make its next action easier to establish. Talk with me about a recurring project handoff.