Champion(s): Matthias Hauser
Project Name: relay.cc
Category: Transactions
*Other: Education, Onboarding
Project Goals:
The high-level goal is to empower users of https://portfolio.metamask.io to learn and interact with the EVM chains through a conversational chat interface. We accomplish the goal by providing (with XMTP team support) a wallet-to-wallet messaging frontend, a wallet-to-AI messaging backend, and an intelligent AI concierge for on-chain actions such as swaps and bridging. Importantly, the AI agent is messaged via ENS name such as chat.metamask.eth, allowing users on any XMTP client to reach the AI agent.
The specific use cases are:
Wallet Chat - Adding universal wallet chat between users to the portfolio app. See the DM feature on https://lenster.xyz, full chat client on https://xmtp.chat, or chat window on https://relay.cc for examples of what could be built.
Realtime Crypto Chatbot - "What are the best performing NFTs in the last week?" This will be done by giving the bot access to helpful and varied APIs and the knowledge of when it should call each of those APIs in order to get certain types of information.
On-Chain Actions Chatbot - “Look up the 3 best stablecoin yields today, and put 10k DAI into each of them” This is the end goal of our grant proposal: instructions for on-chain actions. Our AI agents with knowledge of smart contracts, documentation, and user wallets will be a simpler and faster way to execute complex on-chain actions, as an alternative to typical frontends. The bot will create the transactions for the user to sign with MM to confirm. Although this is a very new idea, we have live proofs of concept for both ENS registration and Uniswap Swaps (seen in Extra Materials section at the end). Forthcoming actions include NFT purchases, bridging, and more. For stability, reliability, and user peace of mind, we will integrate state-of-the-art tx simulations such as Stelo or Fire for a secure visual confirmation step.
At the conclusion, Portfolio users will be able to interact with chat.metamask.eth in order to learn about any topics in web3 and initiate the common types of transactions via conversation. The bot will be available in the portfolio chat window and also on every XMTP client through its ENS name.
Target Milestones:
The bot can help both with knowledge and with transactions, and so we have two deliverables Relay works on in parallel, along with production-grade deployment:
*Deliverable 1: Ask MetaMask Web3 Chatbot
Phase 1: Ask Questions about MetaMask (week 1-3)
Phase 2: Ask Questions about the Web3 Ecosystem (week 4-6)
Phase 3: Ask Questions using Realtime Data Sources (week 7-10)
Future Considerations
*Deliverable 2: On-Chain Actions with AI Concierge
Phase 1: In-Chat Swaps Using Metamask Swaps (week 1-3)
A chatbot will be deployed on the XMTP network that can step a user through executing a Uniswap v3 swap with the configurable Metamask fee
The bot will have a straightforward, direct message-based interface. It will have a specific request/response user flow along the lines of the following:
User: Hello? Bot: Hello, what token would you like to receive? User: UNI Bot: What token would you like to offer? User: ETH Bot: How much ETH would you like to trade for UNI? User: 2 Bot: Calculating... Bot: Great, I found the best trade for you. For 2 ETH, you will receive 607.842 UNI using the Uniswap v3 pair at 0.3% slippage. < Confirm Button >
The confirm button will prompt for a Metamask signature to complete.
Phase 2: Conversational Interface for Metamask Swaps(week 4-6)
For example, it will support user flows along the lines of the following:
User: Hello? Bot: Hello, what token would you like to swap? User: I can't remember the name, it's the most recent Metamask Swap I did. Bot: That was for UNI two days ago. User: Yes, that's it. Bot: How much UNI would you like to swap? You have 207.62 total, with 150 on Ethereum and 57.62 on Optimism. User: Swap all of the UNI on Optimism to the OP token. Bot: Okay sure. Calculating...
Phase 3: Conversational Interface For On-Chain Actions (week 7-10)
Future Considerations
Production deployment:
For both of these deliverables, there is also important backend work:
Bot server improvements to concurrent users
Bot server monitoring
Bot server scaling
Timeline / Path to Production:
Weeks 1-3 → Phase 1 for both deliverables
Weeks 4-6 → Phase 2 for both deliverables
Weeks 7-10 → Phase 3 for both deliverables
About Team:
Sean Brennan — CEO/developer/founder, UC Berkeley ‘13, MakerDAO ‘17-20, Infura ‘20
Achilles Schmelzer — CTO/developer/co-founder, edu: UCLA math, 8 years software engineering background
Andrew Van De Poel — COO/developer/co-founder, edu: CSUSM CS, 10 years software engineering background
Funding Request and Budget Breakdown:
$40,000 USD at the outset to start work on deliverables 1 and 2.
$40,000 USD at the successful delivery of the project goals along with production-grade deployment of the bot server and monitoring.
At the conclusion and based on projected data, we will revisit operating costs and revenue. Note: The cost to run the bot after deployment is proportional to the number of user questions, but should be greatly offset by Swap fees.
Achilles and Andrew will be full-time on this, and Sean will work on the “LLM + EVM” research.
What specific software license does the grantee intend to publish under?
MIT
Other Information / Relevant Links:
Image of the current Uniswap Swap conversation helper: https://gateway.pinata.cloud/ipfs/QmPcAyMmqgfRJdTFtT2VAZQ4yT5jycLiK84YZap7k42Ud7
Image of the current ENS registration helper: https://gateway.pinata.cloud/ipfs/QmbuHuqeP995fZEdneHmuJziB2VHsunjLyvpu4XLqsi26A