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

# Collector Hand

> Autonomous intelligence collector that monitors targets continuously with change detection

## Overview

Collector Hand is an AI-powered intelligence collection engine that monitors any target (company, person, technology, market) 24/7, building a living knowledge graph and detecting changes over time.

**Category:** Data\
**Icon:** 🔍

## What It Does

<Steps>
  <Step title="Define Target">
    Specify what to monitor (company, person, technology, market, or topic)
  </Step>

  <Step title="Continuous Collection">
    Run scheduled sweeps across web sources, news, social media, and documents
  </Step>

  <Step title="Build Knowledge Graph">
    Extract entities and relationships, building a structured intelligence database
  </Step>

  <Step title="Detect Changes">
    Compare each cycle against previous state, identifying new developments
  </Step>

  <Step title="Generate Reports">
    Deliver intelligence reports highlighting key changes and trends
  </Step>
</Steps>

## Configuration

### Target Settings

| Setting              | Description     | Example                                                    |
| -------------------- | --------------- | ---------------------------------------------------------- |
| **Target Subject**   | What to monitor | `Tesla`, `Elon Musk`, `Rust language`, `EV market`         |
| **Focus Area**       | Analysis lens   | `market`, `business`, `competitor`, `person`, `technology` |
| **Collection Depth** | How thorough    | `surface`, `deep`, `exhaustive`                            |

### Collection Schedule

| Setting                   | Options                                 | Use Case               |
| ------------------------- | --------------------------------------- | ---------------------- |
| **Update Frequency**      | `hourly`, `every_6h`, `daily`, `weekly` | How often to collect   |
| **Max Sources Per Cycle** | 10, 30, 50, 100                         | Budget vs thoroughness |

### Intelligence Features

| Setting              | Description                                     |
| -------------------- | ----------------------------------------------- |
| **Alert on Changes** | Publish event when significant changes detected |
| **Track Sentiment**  | Analyze and track sentiment trends over time    |
| **Report Format**    | `markdown`, `json`, `html`                      |

## Activation

### Basic Setup

```bash theme={null}
openfang hand activate collector
```

Configure your target:

```bash theme={null}
openfang hand config collector \
  --set target_subject="Tesla" \
  --set focus_area="business" \
  --set collection_depth="deep" \
  --set update_frequency="daily" \
  --set alert_on_changes="true"
```

### Example Workflow

```
Monitor: Anthropic (company)
Focus: Business intelligence
Frequency: Daily updates
Depth: Deep (full articles + sources)

> Monitor Anthropic for product launches, partnerships, and leadership changes
```

Collector Hand will:

1. Build initial target profile and query set
2. Run daily collection sweeps
3. Extract entities (people, products, events, partnerships)
4. Store in knowledge graph
5. Compare against yesterday's snapshot
6. Generate `collector_report_2026-03-06.md` with changes
7. Alert you if critical changes detected

## How It Works

### 1. Target Initialization

Parses the target subject to determine type:

| Target Type    | Collection Strategy                               |
| -------------- | ------------------------------------------------- |
| **Company**    | Products, leadership, funding, partnerships, news |
| **Person**     | Publications, talks, job changes, social activity |
| **Technology** | Releases, adoption, benchmarks, competitors       |
| **Market**     | Trends, players, reports, regulations             |
| **Competitor** | Product launches, pricing, reviews, hiring        |

Builds initial query set (10-20 queries tailored to target type):

```javascript theme={null}
// Example for company target "Anthropic"
const queries = [
  "Anthropic news",
  "Anthropic product launch",
  "Anthropic partnership",
  "Anthropic funding",
  "Anthropic leadership changes",
  "Anthropic Claude API",
  "Anthropic vs OpenAI",
  "Anthropic latest 2026"
]
```

### 2. Collection Sweep

For each query (up to `max_sources_per_cycle`):

<Steps>
  <Step title="Search">
    `web_search(query)` to find recent results
  </Step>

  <Step title="Fetch">
    `web_fetch(url)` promising articles, press releases, reports
  </Step>

  <Step title="Extract">
    Extract entities: people, companies, products, dates, numbers, events
  </Step>

  <Step title="Tag">
    Attach source URL, timestamp, confidence level, relevance score
  </Step>
</Steps>

**Source quality heuristics:**

| Source Type        | Confidence  | Examples                                     |
| ------------------ | ----------- | -------------------------------------------- |
| Official sources   | High        | Company website, SEC filings, press releases |
| Established media  | Medium-High | Reuters, Bloomberg, TechCrunch               |
| Blog posts, social | Medium      | Personal blogs, Twitter, LinkedIn            |
| Forums, anonymous  | Low         | Reddit, HN comments (without verification)   |

### 3. Knowledge Graph Construction

For each data point:

```javascript theme={null}
// Add entities
knowledge_add_entity({
  type: "person",
  name: "Dario Amodei",
  role: "CEO",
  company: "Anthropic",
  last_seen: "2026-03-06"
})

knowledge_add_entity({
  type: "product",
  name: "Claude 3.5",
  company: "Anthropic",
  category: "LLM",
  launch_date: "2024-06-20"
})

knowledge_add_entity({
  type: "event",
  description: "Series C funding round",
  date: "2025-01-15",
  amount: "$750M",
  participants: ["Google", "Spark Capital"]
})

// Add relationships
knowledge_add_relation({
  from: "Dario Amodei",
  to: "Anthropic",
  type: "founded"
})

knowledge_add_relation({
  from: "Anthropic",
  to: "Claude 3.5",
  type: "launched"
})

knowledge_add_relation({
  from: "Google",
  to: "Anthropic",
  type: "invested_in",
  metadata: {amount: "$300M", date: "2025-01-15"}
})
```

### 4. Change Detection

Compare current snapshot vs `collector_knowledge_base.json`:

**Identified changes:**

| Change Type              | Significance       | Example                                   |
| ------------------------ | ------------------ | ----------------------------------------- |
| **New entities**         | Critical/Important | New executive hire, new product launch    |
| **Changed attributes**   | Critical           | Person changed company, new funding round |
| **New relationships**    | Important          | New partnership, acquisition              |
| **Disappeared entities** | Minor              | No longer mentioned in sources            |

**Scoring changes:**

```javascript theme={null}
function scoreChange(change) {
  // Critical changes (trigger alerts)
  if (change.type === "leadership_change") return "critical"
  if (change.type === "acquisition") return "critical"
  if (change.type === "major_funding") return "critical"
  if (change.type === "product_launch") return "critical"
  
  // Important changes
  if (change.type === "new_partnership") return "important"
  if (change.type === "hiring_surge") return "important"
  if (change.type === "pricing_change") return "important"
  
  // Minor changes
  return "minor"
}
```

If `alert_on_changes = true` and critical changes found:

```javascript theme={null}
event_publish({
  type: "collector_alert",
  target: "Anthropic",
  changes: [
    "New partnership with Google Cloud announced",
    "Dario Amodei interviewed on AI safety"
  ]
})
```

### 5. Sentiment Tracking (Optional)

If `track_sentiment = true`:

For each source, classify:

* **Positive** — Praise, success stories, growth
* **Negative** — Criticism, failures, controversy
* **Neutral** — Factual reporting

Track trend vs previous cycle:

```javascript theme={null}
const sentiment = {
  current: {positive: 60%, neutral: 30%, negative: 10%},
  previous: {positive: 40%, neutral: 50%, negative: 10%},
  trend: "sentiment_improving"
}
```

### 6. Report Generation

**Markdown format:**

```markdown theme={null}
# Intelligence Report: Anthropic
**Date**: 2026-03-06 | **Cycle**: 42 | **Sources Processed**: 28

## Key Changes Since Last Report

### Critical
- **Partnership announced**: Google Cloud strategic partnership for enterprise Claude API
- **Product launch**: Claude 3.5 Sonnet now supports 200K context window

### Important
- Dario Amodei interviewed on Lex Fridman podcast about AI safety
- Hiring surge: 15 new ML research positions opened

## Intelligence Summary

Anthropic continues aggressive expansion in enterprise market with Google Cloud
partnership announced March 5. The deal includes preferential pricing for GCP
customers and integration with Vertex AI. CEO Dario Amodei emphasized safety
and reliability in podcast appearance, positioning against OpenAI's rapid
deployment strategy.

Product development remains focused on context length and safety. Claude 3.5
Sonnet update brings 200K context (vs 100K previously), maintaining quality
at extended lengths. Internal benchmarks show 15% improvement on long-document
tasks.

## Entity Map

| Entity | Type | Status | Confidence |
|--------|------|--------|------------|
| Dario Amodei | Person | CEO, active | High |
| Claude 3.5 Sonnet | Product | Updated | High |
| Google Cloud | Company | New partner | High |
| Vertex AI | Product | Integration | Medium |

## Sources

1. [Anthropic announces Google Cloud partnership](https://anthropic.com/news) — confidence: high — Official press release, March 5
2. [Lex Fridman podcast #412](https://youtube.com/...) — confidence: high — Dario Amodei interview
3. [TechCrunch: Anthropic expands context window](https://techcrunch.com/...) — confidence: medium-high — March 6

## Sentiment Trend
Positive: 60% | Neutral: 30% | Negative: 10%  
Trend: ↑ improving (up from 40% positive last cycle)
```

Saved as: `collector_report_2026-03-06.md`

## Output

| File                            | Description                        |
| ------------------------------- | ---------------------------------- |
| `collector_report_YYYY-MM-DD.*` | Intelligence report for each cycle |
| `collector_knowledge_base.json` | Cumulative knowledge graph         |

## Dashboard Metrics

* **Data Points** — Total intelligence items collected
* **Entities Tracked** — Unique entities in knowledge graph
* **Reports Generated** — Number of reports delivered
* **Last Update** — Timestamp of most recent collection

## Use Cases

### Competitive Intelligence

```bash theme={null}
openfang hand config collector \
  --set target_subject="Competitor Inc" \
  --set focus_area="competitor" \
  --set update_frequency="daily" \
  --set alert_on_changes="true"
```

Monitor:

* Product launches
* Pricing changes
* Customer reviews and sentiment
* Hiring patterns
* Partnerships

### Person Tracking

```bash theme={null}
openfang hand config collector \
  --set target_subject="Sam Altman" \
  --set focus_area="person" \
  --set collection_depth="deep"
```

Track:

* Public appearances
* Interviews and talks
* Social media activity
* Company changes
* Investments

### Market Intelligence

```bash theme={null}
openfang hand config collector \
  --set target_subject="AI agent market" \
  --set focus_area="market" \
  --set collection_depth="exhaustive"
```

Analyze:

* Market size and trends
* New players
* Regulatory changes
* Technology shifts
* Investment flows

## Tips & Best Practices

<Note>
  **For best results:**

  * Start with Daily frequency — Hourly can be noisy for most targets
  * Use Deep collection depth for active targets, Surface for monitoring
  * Enable sentiment tracking for brand monitoring
  * Review knowledge graph monthly to prune outdated entities
  * Combine with Predictor Hand to forecast based on collected trends
</Note>

<Warning>
  Collector Hand only uses publicly available information. Never attempt to collect data behind login walls or from restricted sources.
</Warning>

## Next Steps

<CardGroup cols={2}>
  <Card title="Predictor Hand" icon="crystal-ball" href="/hands/predictor">
    Make predictions based on collected intelligence
  </Card>

  <Card title="Researcher Hand" icon="flask" href="/hands/researcher">
    Deep-dive research on discovered topics
  </Card>
</CardGroup>