> For the complete documentation index, see [llms.txt](https://docs-atlas.rndm.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs-atlas.rndm.io/atlas.md).

# Atlas

<div align="left"><figure><img src="/files/QIhcknVZrqWziOtVH7ka" alt="" width="118"><figcaption></figcaption></figure></div>

Atlas aims to be the first DeFi AI agent yield protocol on Aptos ecosystem.

Instead of you manually:

* Watching pools,
* Checking incentives, swapping adding to pools
* Opening liquidity positions,
* Unwinding them when markets move, and then earning interest
* Whilst losing time..missing out farming incentives.

Atlas does it for you.

### The core idea: agentic yields

Most yield products today are **static**:

* “Deposit into vault X”
* “Vault X always runs strategy Y”
* Share across 100 other depositors
* Fixed allocation rules, fixed pool, fixed behaviour.

Atlas is built around **agents**:

> A yield engine that behaves like a smart agent, not a fixed script.

#### What does “agentic” actually mean here?

* **Observant**\
  Atlas continuously reads on-chain state: vault balances, LP positions, pool info, estimated withdrawals, and incentives.
* **Stateful**\
  Atlas remembers which LP mints belong to which user position (`posId`), what’s allocated, what’s closing, what’s been finalized. So its personal to each user.
* **Task-driven**\
  Instead of you clicking through endless UI steps, Atlas runs jobs:
  * “Allocate fresh deposits into the best stable pool”
  * “Sweep fees DCA (dollar cost average) to the vault every N hours”
  * “Unwind LPs and return USDC when user wants to exit”

It’s doing simpler decisions via LLM and is mostly quantitative, but the architecture is designed so that “which pool/strategy should I use next?” can be driven by more advanced logic over time (models, signals, cross-chain data, etc.). That’s the **agentic** part: the decision engine is flexible and upgradeable, not baked into a single immutable strategy.

***

### Why Atlas feels like a butler

Think of a classic butler:

* He knows where your assets are.
* He keeps track of what’s due when.
* He handles the tedious tasks (collecting, sweeping, unwinding).
* You tell him **intent** (“I’d like my yield served with bonus points please”) and he handles the rest like an  extra cocktail at the pool.

<figure><img src="/files/eIFrSCkgU7wBWiEGfhmu" alt="" width="375"><figcaption></figcaption></figure>

Atlas plays that role for yield:

1. You say:

   > “Here’s USDC; I’d like delta neutral yield with bonus points.”
2. Atlas:
   * Splits USDC into **USDC + APT**,
   * Adds liquidity to the **USDC/APT  pool in Hyperion, and hedge a small % in APY 3X short in Decibel**
   * Tracks all LP mints, positions, and incentives,
   * Periodically collects fees,
   * DCA Apt to maintain very low risk
   * Quotes the **current value** of your position in USDC terms,
   * And when you’re ready to exit, handles the full unwind flow.
   * No lockups, you just have funds back, and no slippage
3. You:
   * Check your **current value** and **APY**,
   * Decide when to close,
   * Withdraw in USDC.
   * see onchain where your positions are, and the exact split

You never need to deal with:

* When to short,
* LP  quirks,
* Which swaps to call,
* How to pick a good time to unwind.
* How to keep collecting fees, and when to maximise decibel hyperion points

Atlas abstracts all of that away as a “yield butler” service.