Mantle Yield Scout: The AI Intelligence Layer Mantle’s Ecosystem Needs
Mantle Yield Scout aggregates every major yield opportunity on Mantle in real time and puts an AI on top of it, so any user can make informed decisions in seconds, not days.
The Problem
A few months ago I wrote a deep-dive on Mantle’s ecosystem, covering the Mantle Vault, Super Portal, Byreal, and Bybit Alpha together as a connected capital system.
It took me three days to write.
Not because the opportunity was complicated. Because the information was completely scattered. Aave V3 data on one dashboard. mETH staking APY on a separate page. Bybit Earn campaigns buried inside Bybit’s product interface. Byreal pool incentives tracked manually. Lendle lending rates from a subgraph. None of it was in one place. None of it was compared side by side. None of it had a risk rating.
I am a researcher. I had the time and the skills to piece it together. Most users do not.
That is the problem. Mantle is one of the most active yield ecosystems in crypto right now, but the information architecture does not match the quality of the protocols
Target user: Any Mantle DeFi participant, from a first-time Bybit user exploring on-chain yield for the first time, to an active LP managing positions across multiple protocols, who wants to make faster, better-informed decisions without spending hours across six different dashboards.
The Solution: Mantle Yield Aggregator (MVP is live🟢)
Mantle Yield Scout is a live yield aggregator for the Mantle ecosystem with an AI assistant built directly on top of the data.
It does three things:
1. Aggregates live yield data across all major Mantle protocols
Every five minutes it pulls real APY, TVL, and pool data from Aave V3 Mantle, Lendle, Beefy vaults, mETH staking, and GeckoTerminal LP pools, normalised into a single ranked feed. One place. Current data. Every protocol.
2. Scores every pool for risk automatically
A custom risk engine classifies each pool into four tiers:
A user does not need to know how to read a subgraph to understand the difference between organic yield and emission farming. The risk tier tells them instantly.
3. Lets any user ask the AI anything in plain English
The AI assistant receives the full live pool snapshot as context on every single message, not a cached summary, the actual current data. A user types a question and gets a real answer built from live numbers.
Examples:
→ “What is the best secure yield for $5,000 right now?” AI filters to Secure-tier pools, returns the top result with protocol name, exact APY, TVL, and step-by-step entry instructions.
How It Interacts With Mantle Onchain
The platform has three jobs running continuously:
- Collect yield data from 7 external sources every 60 seconds
- Score every pool with a 5-metric risk model
- Answer user questions via a streaming AI chatbot backed by live pool data, on-chain reads, and real-time X posts
The entire stack is Next.js 15 running on Vercel. There is no separate backend server. Every API route is a Vercel serverless function.
Get Zayd☯’s stories in your inbox
Join Medium for free to get updates from this writer.
But this is where the project goes beyond a simple dashboard.
For Aave V3 on Mantle, the tool reads data directly from the blockchain. It calls the PoolAddressesProvider contract at 0xba50Cd2A20f6DA35D788639E581bca8d0B5d4D5f to dynamically resolve the live data provider, then batches all reserve reads into a single multicall3 request to the Mantle RPC. This gives real supply APYs, total liquidity, and active market status straight from the contracts, not from a third-party API that can go stale. The Mantle RPC at rpc.mantle.xyz is the primary endpoint, with Ankr and dRPC as automatic fallbacks.
For vaults and liquidity pools, the tool pulls from DefiLlama, Beefy Finance, GeckoTerminal, Lendle and mETH Protocol, simultaneously, deduplicates the data, and enriches each pool with 24-hour trading volume from GeckoTerminal to measure real liquidity depth.
Every pool is scored by a 5-metric risk model: TVL size, how much of the APY is organic base yield versus token emissions, pool type (lending is safer than leveraged LP), liquidity depth measured by volume-to-TVL ratio, and protocol maturity. The result is a consistent risk tier, Secure, Low-Moderate, High, or Very High, applied to every pool on Mantle using the same rules.
The AI Chatbot
The chatbot is powered by Grok-3 and has access to the live pool snapshot on every message. A user can ask “what is the best secure yield for stablecoins right now” and the bot filters by risk tier, sorts by APY, names the protocol, states the current APY and TVL, and tells the user how to enter. It cites real numbers from the aggregator, never fabricated ones.
The bot also searches X (Twitter) using the official API for posts from @Mantle_Official and @Alpha_Bybit from the last 30 days, looking for campaigns, launchpools, vault promotions, and earn incentives. Those results are cached and injected into every conversation, so when a user asks "are there any active campaigns on Mantle", the bot answers with actual recent posts including dates and links.
The knowledge base is extended further through the mantle-skills submodule, a set of structured Mantle DeFi skill documents covering risk evaluation, DeFi operations, and network fundamentals that are loaded into the system prompt on every request.
Every data point comes directly from Mantle’s ecosystem.
Direct RPC integration: The platform uses direct calls to Mantle’s RPC endpoint (rpc.mantle.xyz) via viem to fetch live lending market data, token prices, pool liquidity, and chain status in real time.
Protocol-level data pulls:
- Aave V3 Mantle — lending and borrowing rates, supplied assets, TVL
- Lendle — lending markets via GraphQL subgraph
- Beefy — vault APYs, vault TVL, yield breakdown
- mETH staking — staking APY via Mantle’s native staking API
- GeckoTerminal — LP pool data for Mantle network pools
Risk engine built on Mantle’s yield structure: The four-tier risk classification is designed specifically around how yield is generated on Mantle protocols, distinguishing organic lending interest from MNT emission rewards. This is not a generic risk model; it reflects how Mantle’s incentive programs actually work.
Caching layer: Pool data is cached with a five-minute TTL so every AI response is built on current data without rate-limiting the underlying protocol APIs.
The Vision
The current build is just an MVP that solves the visibility and intelligence problem. Other layers can be integrated like:
On-chain alerts: A user sets a threshold “tell me when mETH APY drops below 4%” or “alert me when a new Bybit × Mantle campaign launches.” The alert layer monitors on-chain data and notifies the user directly. No more manually checking dashboards.
One-click rebalancing: The AI recommends the move. The user approves it in one click. The transaction executes on Mantle. At sub-$0.001 gas, this is only practical on Mantle, not on any other major L2.
Portfolio context: The user connects their wallet once. The AI can now see what they hold, what they have earned, and how their current positions compare to today’s best opportunities. Generic advice becomes personalised intelligence.
Why Mantle
Every piece of this only works because of Mantle’s specific architecture.
The gas costs make frequent rebalancing practical. The Bybit integration means the user base already exists, 82 million CeFi-native users who are one step away from on-chain yield. The protocol depth (Aave V3 crossing $550M TVL in two weeks, the Vault at $150M+ AUM, MI4 at $173M AUM with a 27.9% 2025 return) means the data the AI works with is real and substantial.
The ecosystem is serious. The intelligence layer should match it.
Mantle Yield Scout is that layer.
This article is for educational purposes only and does not constitute financial advice.
