StartAI System
How Start orchestrates Claude agents — independent sessions, shared memory, and resource-aware execution.
Start runs Claude through a background process alongside the app. Each chat session gets its own agent instance, which means you can have multiple agents working in parallel — each on a different task, with its own conversation context.
Loading diagram...
Agent capabilities
Every agent session has access to the same set of tools:
| Capability | What it does |
|---|---|
| File operations | Read, write, and edit files with targeted diffs — no full-file rewrites |
| Terminal | Execute shell commands, capture output |
| Task management | Create, update, and complete tasks on the Kanban board |
| Memory | Save and retrieve long-term notes across sessions |
| Sub-agents | Delegate specialized work to child agents that run in parallel |
Shared context
While each agent has its own conversation history, they share two things:
- Memory — The project's
MEMORY.mdand recent daily logs are injected into every agent's context at session start. This is how agents stay aligned on project conventions and decisions. - Task board — All agents can read and write to the same Kanban board, so work stays coordinated.
Resource awareness
Start monitors your machine's CPU, memory, and thermal state, and throttles agent activity when the system is under load. This prevents agents from overwhelming your machine during parallel execution. See Pressure System.