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

# Core Concepts

> Understand how Adapt works under the hood

Adapt is built on a simple but powerful framework: **Ask**, **Understand**, **Act**. This page explains the core concepts that make Adapt work.

## The Ask-Understand-Act Framework

<Steps>
  <Step title="Ask">
    You ask a question in natural language, just like you would ask a colleague. No SQL, no dashboards, no context-switching.
  </Step>

  <Step title="Understand">
    Adapt investigates across your connected tools and data sources, gathering the information needed to answer your question with cited sources.
  </Step>

  <Step title="Act">
    Based on the insights, Adapt can take action on your behalf—drafting communications, updating records, creating reports, or triggering workflows.
  </Step>
</Steps>

## AI Agents

At the heart of Adapt is an AI agent system. When you ask a question, an agent is dispatched to:

1. **Plan** the steps needed to answer your question
2. **Query** your connected data sources
3. **Analyze** the results and synthesize insights
4. **Execute** any requested actions
5. **Report** back with findings and citations

### Agent Capabilities

Agents can use a variety of tools to accomplish tasks:

| Tool               | Description                                                            |
| ------------------ | ---------------------------------------------------------------------- |
| **Integrations**   | Read and write to HubSpot, Slack, Linear, and other connected services |
| **Data Queries**   | Query Snowflake, Databricks, and other data sources                    |
| **Web Search**     | Search the internet for market research, competitive intel, and more   |
| **Code Execution** | Run Python, JavaScript, or shell scripts in a secure sandbox           |
| **Knowledge Base** | Access your organization's stored knowledge and context                |

### Sub-Agents

For complex tasks, Adapt can spawn sub-agents to work in parallel. For example,
a competitive analysis might involve:

* One sub-agent researching Competitor A
* Another sub-agent researching Competitor B
* A third sub-agent pulling your own metrics for comparison
* The main agent synthesizing all findings into a report

## Adaptive Intelligence

Adapt decides how much intelligence each task needs. Simple, well-defined work runs **Fast**. Hard, open-ended problems run **Deep**, with Adapt's most capable reasoning. You don't pick a mode or a model — Adapt reads the work in front of it and matches its intelligence to the task.

This happens automatically and shifts as the work changes character, even within a single task. There's nothing to configure. It applies to every chat, automation, app, and sub-agent.

The result: routine work stays fast and snappy, while genuinely hard problems get Adapt's deepest reasoning instead of being rushed.

Under the hood, Adapt draws on best-in-class frontier models from Anthropic, Google, and OpenAI, folding in new ones as they arrive. You always get the best of the frontier without ever picking a model.

## Integrations

Integrations connect Adapt to your existing tools and data sources. This allows
Adapt to synthsize insights across sources or take action on behalf of the user.

## Knowledge Base

The knowledge base stores persistent information that Adapt can reference in future conversations:

* **Business context**: Your company's metrics definitions, KPIs, and terminology
* **Procedures**: Standard operating procedures and workflows
* **Facts**: Key information about your business, customers, and products
* **Preferences**: How your team likes reports formatted, which metrics matter most

You can add to the knowledge base explicitly ("Remember that Q4 starts in
October for us") or Adapt will learn over time from your conversations.

For how entries are structured and how to view, add, edit, or remove them,
see the [Knowledge](/features/knowledge) guide.

## Conversations & Threads

### Conversations

Each conversation with Adapt is a thread of messages. Adapt maintains context
throughout the conversation, so you can:

* Ask follow-up questions without repeating context
* Drill down into specific details
* Request different formats or perspectives
* Build on previous analyses

### Thread Visibility

Conversations can be:

* **Private**: Only visible to you
* **Shared**: Visible to anyone in your organization

## Automations

Automations automate recurring queries and actions:

```
Every Monday at 9am:
1. Query last week's sales metrics
2. Compare to targets and previous week
3. Post summary to #sales-team in Slack
```

Tasks can be triggered by:

* **Schedule**: Cron-based timing (daily, weekly, monthly)
* **Events**: Webhooks from external systems

## Sandbox Execution

When Adapt needs to run code for data transformation, visualization, or complex
analysis, it uses a secure sandbox environment:

* **Isolated**: Each conversation runs in its own container
* **Network access**: Can call external APIs as needed
* **Pre-installed tools**: Python, Node.js, pandas, matplotlib, and more
* **Mountable filesystems**: Supports FUSE, so Adapt can mount cloud storage buckets and other sources as local folders (for example with `gcsfuse` or `mount-s3`)

## Next Steps

<CardGroup cols={2}>
  <Card title="Your First Query" icon="play" href="/getting-started/first-query">
    Walk through an example query step-by-step
  </Card>

  <Card title="Integrations" icon="plug" href="/integrations/overview">
    Connect your tools to unlock Adapt's full potential
  </Card>
</CardGroup>