Your AI Team
KingDesk is not a single AI. It's a coordinated team of AI components that work together across your devices and your workflow.
The Three Players
| Component | What it does |
|---|---|
| Claude Code | Your main coding and thinking partner, running inside KingDesk |
| Jarvis | Your assistant — manages todos, sends reports, delegates tasks |
| KingMobile | Your phone companion — approve actions, monitor AI sessions, chat on the go |
They share the same project context. A decision made in KingDesk is visible in KingMobile. A todo created by Claude Code shows up in Jarvis.
KingWork — Teaching Your AI Once
The hardest part of working with AI is re-explaining your preferences every session. KingWork solves this.
KingWork is the system that automatically delivers your project rules and context to Claude Code at the start of every session. Your AI doesn't start blank — it already knows your conventions, constraints, and history.
How it works:
- You write a rule (or your AI discovers one while working)
- It goes into
_Policy/inside your KingFolder - Next session, KingDesk automatically loads it before Claude Code starts
- Your AI already knows the rule — you never repeat yourself
Tell your AI "always use TypeScript strict mode" or "never delete files without moving to trash first." It writes the rule to _Policy/. Every future session, it follows the rule automatically.
What Gets Delivered Every Session
KingDesk injects two layers of context into every Claude Code session:
| Layer | Content |
|---|---|
| Global rules | File naming conventions, coding style, team practices |
| Project context | Your _Policy/ files, environment info from _Infra_Config/, current KingFolder structure |
You can add to either layer at any time. KingDesk picks it up automatically.
Recall — Your AI's Long-Term Memory
Claude Code has a long-term memory called Recall. Past conversations, decisions, and work records are stored and searchable across sessions.
When you say "remember how we decided to handle authentication?" — your AI searches Recall and finds the thread. You don't have to re-explain the context.
What Recall stores:
- Conversation summaries from
__MacFlow/ - Decisions made during past sessions
- Errors encountered and how they were resolved
- Work plans and their outcomes
Recall is searched automatically when your AI senses a past reference. It doesn't announce it — the context just appears in the response.
The Specialist Team
For complex tasks, Claude Code can summon specialist sub-agents:
| Agent | Specialty |
|---|---|
| architect | System design, API interfaces, dependency analysis |
| implementer | Code implementation, bug fixes, performance |
| reviewer | Code review, bug detection, quality checks |
| tester | Test writing, TDD, regression coverage |
| security | Security audit, vulnerability analysis |
| devops | Deployment, CI/CD, environment setup |
Your AI decides when to bring them in. For a simple fix, it works alone. For a full feature, it orchestrates the team automatically.