# AI Agent Network

## Beyond Chat — Autonomous Intelligence

ATLAS doesn't just store knowledge. It deploys **six specialized AI agents** that work autonomously on your behalf.

These agents aren't chatbots. They're workers with specific roles, capabilities, and outputs.

***

## The Agent Squad

### 🔬 Researcher Agent

**Role:** Deep investigation and analysis

The Researcher takes a topic and produces comprehensive analysis:

* Searches across your knowledge graph
* Identifies relevant concepts and content
* Synthesizes findings into structured reports
* Highlights knowledge gaps

**Output:** Research reports with citations to your own knowledge

### 📚 Curator Agent

**Role:** Knowledge organization and quality

The Curator maintains your knowledge graph health:

* Identifies duplicate or overlapping concepts
* Suggests merges and reorganizations
* Scores content quality
* Recommends archival or deletion

**Output:** Curation recommendations and quality scores

### 📝 Reporter Agent

**Role:** Content generation and synthesis

The Reporter transforms knowledge into outputs:

* Generates summaries and digests
* Creates newsletters from recent additions
* Produces topic-specific briefings
* Writes content drafts

**Output:** Ready-to-publish content pieces

### ⚡ Executor Agent

**Role:** Action management and completion

The Executor works through your action queue:

* Prioritizes pending actions
* Researches background for tasks
* Suggests next steps
* Tracks completion status

**Output:** Prioritized task lists with context

### 🧠 Memory Agent

**Role:** Retrieval and recall

The Memory Agent handles intelligent retrieval:

* Answers questions about your knowledge
* Finds relevant prior content
* Surfaces forgotten insights
* Provides contextual recall

**Output:** Precise answers with source citations

### 🛠️ Builder Agent

**Role:** Project execution (BUILDR ZONE)

The Builder turns ideas into implementations:

* Takes project briefs from your knowledge
* Produces code, documents, or artifacts
* Iterates based on feedback
* Delivers completed outputs

**Output:** Working implementations and artifacts

***

## How Agents Work

### Context Injection (Not Tools)

Unlike typical AI assistants that call external tools, ATLAS agents receive **pre-loaded context** from your knowledge graph:

```
Traditional Approach:
Agent → Calls search tool → Waits → Processes results → Responds

ATLAS Approach:
Knowledge Graph → Injects relevant context → Agent analyzes → Responds
```

This means:

* **Faster responses** — No round-trip tool calls
* **Better coherence** — Full context available from start
* **Lower cost** — Efficient context window usage

### Agent Orchestration

Agents can work independently or coordinate:

**Solo Mode:** Single agent handles a task

```
"Research everything I've saved about prediction markets"
→ Researcher Agent → Report
```

**Squad Mode:** Multiple agents collaborate

```
"Prepare a newsletter about my recent knowledge additions"
→ Curator (selects content) → Reporter (writes draft) → Output
```

***

## Agent Capabilities

### Knowledge Tools

| Tool               | Function                           |
| ------------------ | ---------------------------------- |
| `search_knowledge` | Full-text search across your graph |
| `explore_graph`    | Traverse concept relationships     |
| `get_concepts`     | Retrieve specific concepts         |
| `get_actions`      | Access action queue                |
| `get_insights`     | Retrieve insights by topic         |

### Web Tools

| Tool         | Function                    |
| ------------ | --------------------------- |
| `web_search` | Search the broader internet |
| `fetch_url`  | Retrieve web content        |

### File Tools

| Tool             | Function              |
| ---------------- | --------------------- |
| `read_file`      | Access local files    |
| `write_file`     | Create outputs        |
| `list_directory` | Navigate file systems |

***

## Agent Outputs

### Research Report Example

```markdown
# Research: Network Effects in Platform Business Models

## Summary
Based on 23 content items in your knowledge graph spanning 
2022-2024, here's what you've captured about network effects...

## Key Concepts Found
- Network Effects (importance: 0.94)
- Platform Strategy (importance: 0.87)
- Marketplace Dynamics (importance: 0.82)

## Top Insights
1. "Two-sided marketplaces have chicken-egg problems..."
2. "Winner-take-all dynamics depend on switching costs..."

## Knowledge Gaps
- Limited content on regulation impacts
- No content on network effects in AI/ML products

## Suggested Actions
- Research: "Platform regulation effects on network moats"
- Read: "Aggregation Theory" by Ben Thompson
```

***

## Running Agents

### Via CLI

```bash
# Run researcher agent
atlas agent research "topic: prediction markets"

# Run curator agent
atlas agent curate

# Run reporter agent
atlas agent report --format newsletter

# Check agent status
atlas agent status
```

### Via API

```python
POST /api/agents/research
{
  "topic": "prediction markets",
  "depth": "comprehensive"
}
```

### Via Dashboard

Click the "Agents" tab and select which agent to run with your parameters.

***

## Privacy & Control

### Local First

Agents run on your infrastructure. Your knowledge never leaves your machine unless you explicitly expose it.

### Audit Trail

Every agent action is logged:

* What agent ran
* What context was used
* What output was produced
* When it happened

### Control Parameters

You decide:

* Which agents can access which knowledge
* How much context agents receive
* Whether outputs are auto-published or reviewed


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.atlas-ai.org/features/ai-agents.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.