Cellex Agents

Agents on Cellex are autonomous AI agents designed with a backstory, risk tolerance, and a basic foundational algorithm that helps them navigate the Web3 space efficiently. Because they primarily handle real funds in Ethereum, it is important that their foundation is secure and designed for optimal returns on their investments.

The Cellex Agent evaluates newly launched tokens using real-time data from Telegram, Twitter, whitepapers, and market feeds. Based on defined criteria and AI-driven judgment, it decides whether to invest in or divest from a token while operating with a user-defined Ethereum capital allocation

Input Sources: Telegram messages, Twitter feed, whitepapers, web search, and market APIs (OHLCV data).

Autonomous Actions: Buy/sell tokens based on base and dynamic criteria. Monitoring: Agent keeps updating its internal state with new messages and market movements.

Automated Trading: Executes trades based on its evaluation logic.

Components

Data Scraping: Fetches messages/data from TG, Twitter, and other sources. Evaluation: Calculates criteria A (rule-based), B (AI/judgment), and decides action. Execution: Connects to exchanges (or simulated) to execute trades. Crawler: Continuously fetches latest input data and updates agent's evaluation. Market Data API: Fetches OHLCV and other token-specific data.

Evaluation & Decision Logic

A token must meet certain rule-based, quantitative criteria before Cellex considers investing:

  • 𝑳: Liquidity (e.g., 0 if liquidity is 0 and 1 if it’s > $100,000 in first hour)

  • 𝑯: Number of unique holders (e.g., 0 if 0 and 1 if > 500 within 24 hours)

  • 𝑴: Market Cap (e.g., 0 if below $1,000 and 1 if > $10,000,000)

  • 𝒁: Social buzz (sigmoid function of messages + tweets + mentions per day)

  • π‘ͺ: Smart contract safety, 1 – probability of red flags (rug-pull functions, blocking sell functions, etc.)

  • 𝑺: Sentiment Score from TG & Twitter (NLP sentiment analysis, range 0 to 1, 0.5 means neutral)

  • 𝑸: Whitepaper quality (LLM/AI score combining relevance, completeness and uniqueness)

  • 𝑡: Token novelty (LLM/AI output, cosine similarity with known scams/legit projects)

  • 𝑾: Web search findings (scam reports, listing status, partnerships, etc.)

  • 𝑢: Organic community engagement (number and diversity of contributors)

  • πœƒπ‘π‘’π‘¦: Buy threshold (e.g. 0.7)

  • πœƒπ‘ π‘’π‘™π‘™: Sell threshold (e.g. 0.4)

Formula:

π‘Œ=𝑀1.𝐿+𝑀2.𝐻+𝑀3.𝑀+𝑀4.𝑍+𝑀5.𝐢+𝑀6.𝑆+𝑀7.𝑄+𝑀8.𝑁+𝑀9.π‘Š+𝑀10.𝑂

All variables will be normalized between 0 and 1 If certain variable is unavailable, its weight can be turned to 0 to ignore it during calculation. The sum of all weights 𝑀1+𝑀2+⋯𝑀𝑛=1

Decision:

If π‘Œ>πœƒπ‘π‘’π‘¦, then Cellex agent executes β€œbuy” function; if π‘Œ<πœƒπ‘ π‘’π‘™π‘™, then the agent executes β€œsell” function; if πœƒπ‘ π‘’π‘™π‘™<π‘Œ<πœƒπ‘π‘’π‘¦, then the tokens are in hold position.

Trade Volume

  • 𝑬: available capital in ETH authorized to the Cellex agent

  • π‘³π’Žπ’‚π’™: portion of available funds that can be invested in an individual token

  • 𝑹: risk tolerance (between 0 and 1) per token

  • 𝑿: the amount in ETH. When selling, 𝑋 will be positive; when buying, it will be negative

Example:

If a new token β€œCellex Token” is launched and its data is as:

  • Liquidity reaches $25,000 in first hour, 𝐿=0.25 (i.e., 25000/100000)

  • 400 unique holders in 24h, 𝐻=0.8 (i.e., 400/500)

  • $500,000 Market Cap, 𝑀=0.049 (i.e., (500,000-1,000)/10,000,000)

  • 𝑍=0 (i.e., data not available)

  • Smart contract safety, 𝐢=0.7 (i.e., 0.3 chance of malicious code in contract)

  • Sentiment Score from TG & Twitter, 𝑆=0.67 (mildly positive)

  • 𝑄=0 (i.e., data not available)

  • 𝑁=0 (i.e., data not available)

  • π‘Š=0 (i.e., data not available)

  • 𝑂=0 (i.e., data not available)

  • Buy threshold, πœƒπ‘π‘’π‘¦=0.45 (assuming market is very bullish)

  • Sell threshold, πœƒπ‘ π‘’π‘™π‘™=0.2

If all available variables have equal weightage, then:

𝑀1=𝑀2=𝑀3=𝑀5=𝑀6=15=0.2 𝑀4=𝑀7=𝑀8=𝑀9=𝑀10=0

Therefore, the decision to buy or cell will be: π‘Œ=0.2Γ—0.25+0.2Γ—0.8+0.2Γ—0.049+0.2Γ—0.7+0.2Γ—0.6=0.4798 Since π‘Œ>πœƒπ‘π‘’π‘¦, therefore the Cellex Agent will decide to BUY Cellex Token. To calculate the amount against which the token will be bought:

  • 𝑬: 1.5ETH

  • π‘³π’Žπ’‚π’™: 0.05 (only 5% of funds can be utilized in buying a token)

  • 𝑹: 0.5 (the risk tolerance will be 1 if π‘Œ=1)

𝑋=1.5Γ—0.05Γ—0.5

Conclusion:

Last updated