# Agent Mode

Agent mode deploys an autonomous AI trading agent that continuously monitors markets, evaluates tokens, makes decisions, and executes trades on your behalf. Set your strategy once, and the agent runs 24/7.

***

## What an Agent Does

An AI agent is an autonomous trading system that:

1. **Observes** — Continuously monitors market signals, prices, and on-chain data for tokens on your watchlist
2. **Orients** — Processes signals through your configured strategy and risk parameters
3. **Decides** — Uses AI reasoning to evaluate whether to buy, sell, or hold
4. **Acts** — Executes trades via Jupiter when decisions are made
5. **Learns** — Records decisions and outcomes to improve future decisions

***

## Before You Start: Key Concepts

### Paper Mode vs Live Mode

Agents start in **Paper Mode** — they simulate decisions but don't execute real trades. This lets you test your agent's strategy risk-free.

Switch to **Live Mode** when you're confident in the strategy. In Live Mode, real trades execute from your wallet.

### OODA Loop

The agent follows the OODA (Observe-Orient-Decide-Act) loop:

* **Observe**: Read current market data — price, volume, signals
* **Orient**: Apply your strategy rules and agent memory
* **Decide**: AI evaluates the situation and makes a decision
* **Act**: Execute buy/sell/hold if in Live Mode

### Agent Memory

Agents maintain a memory system that improves over time:

* **Recent trades**: Past decisions and outcomes
* **Win/loss streaks**: Consecutive wins or losses
* **Market regime**: Bull, bear, or sideways market detection
* **Learned patterns**: Patterns the agent has identified from its trades

***

## Creating an Agent

### Step-by-Step

1. **Send `/agent`** — Opens the agent management panel.
2. **Select "Create Agent"** — Starts the configuration flow.
3. **Choose a strategy template** — Select from available templates:
   * **Momentum** — Follows price trends
   * **Breakout** — Buys when price breaks resistance
   * **Scalper** — Quick small-profit trades
   * **Sniper** — Targets new token launches
   * **Mean Reversion** — Buys oversold, sells overbought
4. **Name your agent** — Uppercase letters, numbers, and hyphens only. Example: `NEXUS-7`
5. **Set your trading personality**:
   * **CAUTIOUS** — Small positions (5% of portfolio), tight stops (-10%), lower risk
   * **BALANCED** — Moderate positions (10% of portfolio), normal stops (-20%)
   * **AGGRESSIVE** — Larger positions (20% of portfolio), wider stops (-30%)
6. **Confirm** — ARCHONE creates your agent and shows its dashboard.

***

## Strategy Templates

### Momentum

Follows tokens showing strong upward momentum — price increases combined with volume spikes.

**Best for:** Bull markets, trending tokens

| Parameter        | Value                                                   |
| ---------------- | ------------------------------------------------------- |
| Entry conditions | 5% price increase + 2x volume spike + safety score > 60 |
| Position size    | 5% of portfolio                                         |
| Stop loss        | -15%                                                    |
| Take profit      | +30%                                                    |
| Max hold         | 60 minutes                                              |

### Breakout

Buys tokens breaking through resistance levels with volume confirmation.

**Best for:** Volatile markets, tokens with clear support/resistance

| Parameter        | Value                                           |
| ---------------- | ----------------------------------------------- |
| Entry conditions | 10% price break + 3x volume + safety score > 50 |
| Position size    | 10% of portfolio                                |
| Stop loss        | -10%                                            |
| Take profit      | +50%                                            |
| Max hold         | 120 minutes                                     |

### Scalper

Executes quick trades on small price movements. High frequency, small targets.

**Best for:** Volatile tokens, market maker ranges

| Parameter        | Value                                    |
| ---------------- | ---------------------------------------- |
| Entry conditions | 2% price change + 1.5x volume + RSI < 30 |
| Position size    | 3% of portfolio                          |
| Stop loss        | -5%                                      |
| Take profit      | +10%                                     |
| Max hold         | 15 minutes                               |
| Max daily trades | 20                                       |

### Sniper

Targets new token launches with high potential. Waits for liquidity and safety checks.

**Best for:** Early-stage tokens, new meme coins

| Parameter        | Value                               |
| ---------------- | ----------------------------------- |
| Entry conditions | Safety score > 70 + 5x volume spike |
| Position size    | 2% of portfolio                     |
| Stop loss        | -25%                                |
| Take profit      | +100%                               |
| Max hold         | 240 minutes                         |
| Min liquidity    | $10,000                             |

### Mean Reversion

Buys tokens that have dropped significantly, expecting a rebound.

**Best for:** Oversold tokens, range-bound markets

| Parameter        | Value                           |
| ---------------- | ------------------------------- |
| Entry conditions | RSI < 30 + price below baseline |
| Position size    | 8% of portfolio                 |
| Stop loss        | -10%                            |
| Take profit      | +20%                            |
| Max hold         | 90 minutes                      |

***

## Agent Dashboard

After creating an agent, you see its dashboard showing:

* **Status**: Active, Paused, or Stopped
* **Mode**: Paper (simulated) or Live (real trades)
* **P\&L**: Total profit/loss in SOL
* **Win Rate**: Percentage of winning trades
* **Total Trades**: Number of trades executed
* **Current Streak**: Consecutive wins or losses
* **Last Signal**: Time of last market evaluation

***

## Managing Your Agent

### Pause / Resume

Tap **Pause** to temporarily stop the agent's cycles. Pausing:

* Halts all market monitoring and decisions
* No trades execute (paper or live)
* Resume by tapping **Resume**

Use pause when you want to temporarily stop trading without deleting the agent.

### View Statistics

Tap **Stats** to see detailed agent performance:

* Trade history
* Average hold time
* Best and worst trades
* Learning progress

### Run a Manual Cycle

Tap **Run** to force a manual evaluation cycle. The agent immediately:

* Fetches current market data
* Evaluates all tokens
* Makes decisions
* Executes (if in Live Mode)

### Switch to Live Mode

When ready to trade with real money:

1. Tap **Mode** on the agent dashboard
2. Review the confirmation message
3. Confirm — this switches from Paper to Live

**In Live Mode, real trades execute from your wallet.** Make sure you have sufficient SOL balance.

***

## Agent Decision Process

This section details how the agent makes decisions during each cycle.

### Step 1 — Fetch Market Data

For each token on the watchlist, the agent fetches:

* Current price from Jupiter
* Safety scan results from RugCheck
* Volume and price change signals from DexScreener
* Any open position details

### Step 2 — Evaluate Signals

The agent compiles signals into an overall score:

* Price momentum (positive = bullish, negative = bearish)
* Volume anomalies (spike = unusual activity)
* Buy/sell pressure (more buys = bullish)
* Holder concentration (high selling = bearish)
* Liquidity (higher = safer)

### Step 3 — Apply Strategy Rules

Each template has entry/exit rules. The agent checks:

* Does the current market data match the entry conditions?
* Does a current position match exit conditions?
* Are risk limits respected?

### Step 4 — Make AI Decision

With strategy rules as guidelines, the agent calls AI to make a final decision:

* BUY: Entry conditions met, AI confident
* SELL: Exit conditions met or risk detected
* HOLD: Conditions not met, waiting
* SKIP: Insufficient data or safety concerns

### Step 5 — Execute (Live Mode Only)

If in Live Mode and a decision is made:

* Agent calculates position size based on your personality setting
* Executes via Jupiter
* Records the decision in memory
* Notifies you

***

## Self-Evaluation

Agents continuously self-evaluate their performance and may adjust behavior.

### What Gets Evaluated

After each cycle, the agent reviews:

* Win rate trend (improving or declining)
* Consecutive losses (risk of drawdown)
* Recent P\&L (positive or negative)
* Consistency (stable or volatile returns)

### Possible Recommendations

| Recommendation   | Meaning                             | Action                          |
| ---------------- | ----------------------------------- | ------------------------------- |
| **CONTINUE**     | Performance is healthy              | Keep running as-is              |
| **REDUCE\_SIZE** | Win rate declining                  | Advisor to reduce position size |
| **PAUSE**        | Significant performance degradation | Agent pauses itself             |

**Note:** PAUSE is a safety measure. The agent pauses itself if performance degrades significantly. You can manually resume.

***

## Agent Memory System

Agents remember their past to improve decisions.

### What Gets Recorded

Every decision is stored with:

* Token evaluated
* Decision made (buy/sell/hold/skip)
* Confidence level
* Reasoning
* Market context at the time
* Outcome (win/loss when resolved)

### How Memory Is Used

When evaluating a token:

* Agent retrieves past decisions on the same token
* Considers patterns from similar market conditions
* Learns what worked and what didn't

### Memory Limitations

* Memory is agent-specific — agents don't share memories
* Old memories have lower importance
* Patterns require multiple observations before being considered reliable

***

## Requirements to Run

For the agent to function:

| Requirement               | Why It Matters                          |
| ------------------------- | --------------------------------------- |
| **SOL balance**           | Needed for trades and fees              |
| **Agent status = Active** | Paused agents don't run cycles          |
| **AI credits**            | AI reasoning consumes credits per cycle |
| **Market data available** | If all APIs fail, agent skips cycle     |

***

## Limitations

* Agents don't guarantee profit — they automate your strategy, not guarantee returns
* AI decisions can be wrong — always monitor in the beginning
* Paper Mode results don't predict Live Mode results
* Agents only trade tokens on their watchlist
* Network issues can cause missed cycles


---

# 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.archone.trade/core-features/agent.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.