# Launching an Agent

Deploying an agent takes four steps. The entire process is non-custodial, you retain full ownership of every wallet.

Step 1 — Identity

Could you give your agent a name and write its personality and strategy in plain language? This prompt becomes the agent's system instruction: it defines how it thinks, what it prioritizes, and how it reacts to market conditions.

**Risk profiles:**

| Profile      | Behavior                                           |
| ------------ | -------------------------------------------------- |
| Conservative | Stable pairs, tight ranges, minimal drawdown       |
| Moderate     | Balanced between yield and volatility              |
| Aggressive   | High-yield pools, wider ranges, active rebalancing |

Step 2 — Equip Skills

Skills are onchain modules that unlock specific actions for your agent. They are purchased once with [**RFL tokens**](https://docs.reflect.now/token-details) and permanently attached to the agent wallet.

See the Skills section for the full catalogue.

Step 3 — Fund the Agent

Transfer ETH to the agent's dedicated wallet. The agent uses ETH as gas and as capital for swaps and liquidity operations.

**Minimum recommended:** $10–20 worth of ETH for meaningful autonomous operation.\
**Gas reserve enforced:** The agent always preserves a minimum of $2 worth of ETH to cover future transactions. It will never spend itself into an unusable state.

Step 4 — Review & Launch

Confirm the configuration and start the agent. Once started, it enters an autonomous decision loop that checks market conditions, forms a plan, and executes onchain actions.