> ## Documentation Index
> Fetch the complete documentation index at: https://docs.embedder.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Core concepts

> Understand how Embedder chooses tools, manages context, controls execution, and carries work across sessions in firmware and hardware projects.

Core concepts explain how Embedder works with your project. Start with the common workflow, then learn how modes, approvals, and session context affect each turn.

## Understand the operating model

Embedder combines four controls during a task:

1. Project context supplies source files, `EMBEDDER.md`, hardware selections, schematics, documents, skills, and saved memory.
2. The agent mode determines whether the session plans, implements, or uses debugging tools.
3. The approval policy determines which available tool calls wait for confirmation.
4. The session keeps the conversation, queued prompts, checkpoints, and compressed context.

Modes and approvals are separate. For example, Act mode can edit files, but an edit still waits when the active approval policy requires confirmation.

<CardGroup cols={2}>
  <Card title="Embedded development workflows" icon="route" href="/core-concepts/common-workflows">
    Set up a project, plan changes, build, test, flash, observe, and debug.
  </Card>

  <Card title="Agent modes" icon="toggle-on" href="/core-concepts/modes">
    Choose Act, Plan, or Debug based on the work and evidence you need.
  </Card>

  <Card title="Sessions and context" icon="comments" href="/core-concepts/sessions">
    Manage chat tabs, history, queued prompts, checkpoints, forks, and sharing.
  </Card>

  <Card title="Tool approvals" icon="shield-check" href="/core-concepts/approvals">
    Control confirmations and configure persistent command and path rules.
  </Card>
</CardGroup>

## Make work reliable and reusable

Use these concepts after you understand the basic operating model:

<CardGroup cols={2}>
  <Card title="Get reliable results" icon="circle-check" href="/core-concepts/best-practices">
    Write bounded prompts, attach precise context, and require observable evidence.
  </Card>

  <Card title="Skills" icon="puzzle-piece" href="/core-concepts/skills">
    Load reusable instructions and install compatible skill packages.
  </Card>

  <Card title="Project memory" icon="brain" href="/core-concepts/memory">
    Save durable project knowledge and recall it in later sessions.
  </Card>

  <Card title="Subagents and agent teams" icon="users" href="/core-concepts/agent-teams">
    Delegate isolated research or parallelize work with explicit boundaries.
  </Card>

  <Card title="Goals and bounded loops" icon="repeat" href="/core-concepts/goals-and-loops">
    Continue toward an objective or run reviewed work on a fixed cadence.
  </Card>
</CardGroup>

<Tip>
  If this is your first project, complete the [Quickstart](/quickstart) before reading the concept pages in detail.
</Tip>
