Integration Grants Application '25: Kaja Autonomous Treasury

Name of the project: Kaja Autonomous Treasury

Problem statement
In the Conflux ecosystem, on-chain trading lacks AI, social and financial engagement tools, especially for meme coins. User flow often relies on deployers to launch and promote a coin limiting a community trust, and raising a risk of insiders. Another ecosystem factor is an absence of a low risk investment strategies, where users could benefit from profitable trading activity while not taking an asset price risk. This creates a gap where users lack tools to collaboratively influence trading strategies while getting rewards for successful ones.

Proposed solution
Kaja addresses this by introducing an Autonomous AI Treasury that manages its own funds on the Conflux blockchain, evaluates community-driven trade proposals (shills), executes trades independently and tweets about here activity on X. This fosters a transparent, decentralized, and interactive trading experience.
Every interaction, transaction, and trade outcome is publicly recorded on the Conflux blockchain, ensuring full transparency and accountability.

Key Concepts

  • Community driven suggestions (Shills):
    Users interact with Kaja via a web chat interface, proposing token trades by providing a DEX pool link or token address along with supporting arguments.
  • Autonomous Decision-Making:
    Once a shill attempt is submitted, Kaja autonomously decides whether to seek more details (Discussion), decline the proposal (Reject), or approve and execute the trade (Approve).

Treasury Model & Financial Sustainability

The treasury starts with an initial capital of $4,000, allocated as follows:

  • Low-Risk Portion: $2,000, dedicated to yield farming strategies to generate stable, low-risk revenue.
  • High-Risk Portion: $2,000, used for speculative trading of meme coins based on user shills, acknowledging the potential for high returns and high losses.

This 50-50 split ensures a balanced approach, with half the treasury focused on stability and the other half on growth, aligning with the requirement to allocate 50% to lower-risk strategies.

Revenue Streams and Operations

  1. Shill Fees:
    • Users pay $1 per shill attempt to suggest meme coins for Kaja to consider, with an expected of 100+ shills per month (during the active marketing campaign), generating at least $100 in monthly revenue.
  2. Yield Farming Strategies:
    • The low-risk portion ($2,000 initially) is allocated to yield farming on Conflux-based DEXs such as Swappi and GraFun. Possible strategies include:
      • Liquid Staking: Staking Conflux’s native token, CFX, to earn staking rewards, with an assumed annual yield of approximately 5-8% based on general DeFi practices (e.g., Conflux Network mentions built-in staking at 4%, but we assume 7% for modeling).
      • Bluechip DEX Liquidity Positions: Providing liquidity in stablecoin pools (e.g., USDT/CFX) and CFX/stables pools, with an estimated APY of 5-10% based on current market conditions.
  3. Profits from Speculative Meme Coin Trading:
    • The high-risk portion is used for trading meme coins suggested by users, with the AI evaluating each shill based on market data and sentiment. The trading strategy involves:
      • Buying tokens using the high-risk portion’s capital.
      • Selling tokens at a profit or loss, with net profits calculated monthly.
      • Profit distribution: 50% of trading profits are retained in the high-risk portion to fuel further trading, while 50% is set aside as rewards to incentivize community participation. For example, if a trade yields a $100 profit, $50 is added back to the high-risk portion, and $50 is distributed as rewards.
    • Losses from unsuccessful trades directly reduce the high-risk portion, highlighting the need for robust risk management.
  4. Donations:
    • The proposal mentions donations as an additional revenue stream. These are treated similarly to shill fees, added to the high-risk portion to increase trading capital. While unpredictable, donations can provide a buffer during periods of trading losses.

Automatic Adjustments

The shill fees provide a consistent revenue stream, added to the high-risk portion to enhance trading capacity. The low-risk portion’s yield adds stability, while trading profits (if positive) can accelerate growth, with 50% reinvested.

  • If total treasury T < \$3,000 or high-risk portion H < \$1,600, reallocate to 70% low-risk, 30% high-risk to reduce exposure.
  • If T > \$5,000 and H>$2,000H > $2,000H > \$2,000, reallocate to 30% low-risk, 70% high-risk for growth.
  • For values between $3,000 and $5,000, maintain 50-50.

This dynamic reallocation ensures that as the treasury grows, it can take on more risk for higher returns, but if it shrinks, it reduces exposure to protect remaining capital. An unexpected detail is the integration of shill fees as a direct funding mechanism for high-risk trading, which is uncommon in traditional finance but aligns with Kaja’s community-driven model.

AI Decision-Making & Execution Strategy

Kaja agent processes the following data to operate the treasury:

- Market Data: Real-time prices, liquidity, and volatility from Conflux DEXs like Swappi and GraFun, and general market data from Dexscreener and CoinGecko (for bluechips).
- Social Sentiment: Analysis of user shills and social media trends (from X) to evaluate meme coin potential, including engagements metrics within smart accounts (from Elfa and Yappers).
- User Reputation: Scores based on past shill success to prioritize reliable suggestions, implemented through a reputation system to counter manipulation.
- Risk Metrics: Volatility and liquidity thresholds to filter out risky trades, such as position sizing (max 5-10% of treasury per trade) and stop-loss mechanisms.
- Yield Farming Data: Current APYs from Conflux DEXs and native staking.
- Reliability: Kaja uses minimum two sources of data to secure itself from service disruptions (CoinGecko and Dexscreener in parallel). The same approach is used for RPCs and LLM APIs (using ChatGPT 4o and DeepSeek R1).

Alignment with the Conflux Network
By integrating with Conflux-based DEXs like Swappi and meme coin launchpad GraFun, Kaja enhances the ecosystem’s DeFi capabilities with actively trading new coins and blue chips, interacting with users, sharing profits and gifting rewards to the most active community members.

Benefit to the Conflux Ecosystem

  1. Boosting on-chain activity with Kaja trading, investing and social activity.
  2. User engagement to interact with Kaja’s Treasury (by shilling their own meme or suggesting her an investment ideas without putting user’s assets at a risk).
  3. Attract developers to build their trading agents to interact with Kaja’s Treasury.
  4. Asset Growth: The initial treasury and subsequent profits will increase assets locked in the ecosystem, enhancing liquidity and economic activity.

Links to Project:
KAJA: https://kaja.app
X/Twitter: https://x.com/kaja_ai
Docs: https://zpoken.notion.site/KAJA-Docs-1a79506ab148807ca597dc4f4196d828?pvs=4
Github: https://github.com/ZpokenWeb3/ai-agent-rust-backend
MPC Wallet: https://github.com/ZpokenWeb3/mpc-tss-lib

Conflux eSpace Grant Recipient Wallet Address:
0xDF051FBD601672efbd3C7C0f4eF88C8971dA4A1F

Competitive Edge
Unlike traditional trading bots or platforms reliant on user funds, Kaja operates with its own treasury, ensuring independence and reducing user risk. Its community-driven shill system, paired with transparent blockchain records, sets it apart from centralized or opaque trading solutions.

Are you an incorporated startup?
Yes, Kaja is developed by Zpoken, an Estonian-based company specializing in protocol research and blockchain engineering.

Technical Proposal

Functional Overview
Kaja operates as an AI-driven Treasury system with:

  • Chat Interface: Web app for user shill submissions.
  • AI Decision Engine: Evaluates proposals based on market data and user input.
  • Treasury Management: Executes trades using its own funds via Conflux DEXs.
  • State Tracking: Records and displays portfolio metrics on-chain.
  • Leaderboard: tracks users and rank them for multiple activity metrics.

Core Aspects

  • Treasury service: Handle treasury funds, trade execution, and profit distribution.
  • DEX Integration: APIs for Swappi and GraFun to parse pools and execute swaps.
  • Off-chain Data: 3rd party APIs (web search, dex tools, coingecko).
  • X/Twitter: posting and replying.

Legal/Licensing Aspects
Open-source under MIT License, ensuring community access and auditability.

Marketing Strategy for Kaja
To drive user engagement and platform adoption, we will execute a targeted marketing campaign leveraging on-chain incentives and social engagement. Our $4,000 budget will be allocated across three key initiatives:
Advisory Challenge – Users who successfully convince Kaja to buy a token that results in the most profitable trade will be rewarded. The top three most profitable calls will receive prizes, encouraging strategic participation.
Shiller/Degen Leaderboard – The most active users who repeatedly persuade Kaja to buy tokens will be ranked, with top participants receiving rewards. This campaign fosters engagement and continuous interaction with the AI agent.
Partners Campaign – A community-driven initiative where random users who takes part in mutual AMA sessions will win a prize.
KOLs Campaign – A campaign on Twitter / X (paid marketing) + social campaign (quiz) on Intract.xyz for the community to engage. Total campaign cost would be $1,000

These campaigns will be complemented by organic social engagement, viral marketing via X (Twitter),

Total Budget
We are requesting a total grant of $8,000 to cover initial treasury funding and marketing efforts. Justification:

  • Initial Treasury (seed funds for Kaja’s trading (e.g., meme coins, blue chips on Conflux): $4,000;
  • Marketing/Rewards Campaigns: $4,000.
    Additional funding (e.g., development costs) provided by Zpoken.

Timeline and Milestones

Month 1: Setup and Integration
Deliverable:

  • Setup Kaja on Conflux eSpace;
  • Integrate with Swappi and GraFun;
  • Fund initial Treasury with $4k in assets.
    Funding: $4,000

Month 2-5: Launch and Marketing
Deliverable: Kaja live with its treasury;
2-Week Challenges (for 4 consecutive months):

  • Advisory Challenge Prize: $100/$50/$25 (1st/2nd/3rd place);
  • Shiller/Degen Leaderboard: $100/$50/$25 (1st/2nd/3rd place);
  • Partners Campaign: $25 (for 4 random users) based on quantity of partner campaigns.
  • KOLs Campaign on Twitter / X + social campaign (quiz) on Intract.xyz for the community to engage. Total campaign cost would be $1,000.
    Launch & Marketing Funding: $4,000

Team
Kaja team comprises experienced professionals from Zpoken. Our team brings together a diverse skill set with expertise in cryptography, blockchain development, product design, and user experience.

Check out our team here: https://www.notion.so/zpoken/Zpoken-Cryptography-Protocol-Research-Team-106533a95ec64c1b903cbe918efbab77

Terms of Use
We agree to all of the following terms of use in applying to a Conflux Ecosystem Grant:

  • we have read and understood the Conflux Grants Ecosystem Overview.
  • we have read about and understood that the Conflux Technical Grants are subject to a No-Sale rule.
  • we agree to provide KYC information to the Conflux Foundation for the sake of overall ecosystem security.
  • we understand that we will be required to follow public grant reporting requirements.

Thank you for submitting your grant application for Kaja Autonomous Treasury. We appreciate the innovative approach you are taking to integrate AI-driven trading with community engagement on the Conflux blockchain. The concept of an autonomous treasury combined with a shill-driven trading mechanism is compelling.

After reviewing your proposal, we would like further clarification on key areas before moving forward with the evaluation process. Your responses will help us better assess the feasibility and impact of Kaja within the Conflux ecosystem.

  1. Treasury Model & Financial Sustainability
    • The proposal outlines an initial $4,000 treasury. Could you clarify how Kaja will sustain its operations beyond this seed funding?
    • If the treasury is depleted due to unsuccessful trades, will additional funding be required, or does the AI have a self-sustaining reinvestment strategy?
    • What mechanisms are in place to manage risk, such as stop-loss orders, position sizing, or portfolio rebalancing?

  2. AI Decision-Making & Execution Strategy
    • How does Kaja evaluate which trades to execute? Does it rely purely on community sentiment, or does it incorporate market analysis (e.g., technical indicators, liquidity depth, volatility)?
    • Have any backtests been conducted on the AI’s trading strategy? If so, what success rates or key metrics have been observed?
    • Given that the AI relies on external APIs (DEX tools, CoinGecko), how does the system handle service disruptions or inaccurate data feeds?

  3. Security & Smart Contract Risks
    • Have the treasury management contracts undergone an independent security audit? If not, are there plans for an audit before launch or is an audit planned for later?
    • How is treasury custody managed? Since you mention an MPC wallet, what safeguards are in place to prevent unauthorized access or potential exploitation?
    • Could Kaja be vulnerable to market manipulation via coordinated shilling attempts (e.g., a group of users artificially promoting illiquid assets)? How do you intend to mitigate this risk?

  4. Growth & Engagement on Twitter/X
    • Kaja will be posting trades and engaging on Twitter—how will you ensure its tweets gain visibility beyond the existing Conflux community?
    • What content strategy will Kaja follow? (E.g., meme engagement, alpha calls, AI-driven market insights, interactive polls?)
    • Are there any plans for collaborations with existing meme coin communities or influencers to drive visibility?
    • How will you measure and optimize engagement metrics (e.g., likes, retweets, mentions) to track growth?
    • Given that AI-generated tweets often struggle with authenticity and relatability, how will Kaja’s Twitter activity be curated to ensure it sounds human-like and engaging rather than robotic?
    • Are there any paid marketing campaigns planned for Twitter/X ads, or will growth be entirely organic?

  5. Team & Long-Term Vision
    • While Zpoken appears to have expertise in cryptography and blockchain engineering, could you highlight specific team members with experience in AI-based trading systems?
    • Your proposal outlines a two-month roadmap. Could you provide additional details on the long-term plans post-launch? For example, how do you see Kaja evolving in the next 6–12 months? We do want the grants program to ba a great stepping stone for sustainable projects.

We appreciate your time and effort in responding to these questions. Strengthening these aspects of your proposal will help us better assess its viability and alignment with the Conflux ecosystem.

Looking forward to your responses.

Thank you, Christian, for your feedback, it is very valuable for us to make Kaja better and more mature!

1. Treasury Model & Financial Sustainability

The treasury starts with an initial capital of $4,000, allocated as follows:

  • Low-Risk Portion: $2,000, dedicated to yield farming strategies to generate stable, low-risk revenue.
  • High-Risk Portion: $2,000, used for speculative trading of meme coins based on user shills, acknowledging the potential for high returns and high losses.

This 50-50 split ensures a balanced approach, with half the treasury focused on stability and the other half on growth, aligning with the requirement to allocate 50% to lower-risk strategies.

Revenue Streams and Operations

  1. Shill Fees:
    • Users pay $1 per shill attempt to suggest meme coins for Kaja to consider, with an expected of 100+ shills per month (during the active marketing campaign), generating at least $100 in monthly revenue.
  2. Yield Farming Strategies:
    • The low-risk portion ($2,000 initially) is allocated to yield farming on Conflux-based DEXs such as Swappi and GraFun. Possible strategies include:
      • Liquid Staking: Staking Conflux’s native token, CFX, to earn staking rewards, with an assumed annual yield of approximately 5-8% based on general DeFi practices (e.g., Conflux Network mentions built-in staking at 4%, but we assume 7% for modeling).
      • Bluechip DEX Liquidity Positions: Providing liquidity in stablecoin pools (e.g., USDT/CFX) and CFX/stables pools, with an estimated APY of 5-10% based on current market conditions.
  3. Profits from Speculative Meme Coin Trading:
    • The high-risk portion is used for trading meme coins suggested by users, with the AI evaluating each shill based on market data and sentiment. The trading strategy involves:
      • Buying tokens using the high-risk portion’s capital.
      • Selling tokens at a profit or loss, with net profits calculated monthly.
      • Profit distribution: 50% of trading profits are retained in the high-risk portion to fuel further trading, while 50% is set aside as rewards to incentivize community participation. For example, if a trade yields a $100 profit, $50 is added back to the high-risk portion, and $50 is distributed as rewards.
    • Losses from unsuccessful trades directly reduce the high-risk portion, highlighting the need for robust risk management.
  4. Donations:
    • The proposal mentions donations as an additional revenue stream. These are treated similarly to shill fees, added to the high-risk portion to increase trading capital. While unpredictable, donations can provide a buffer during periods of trading losses.

Automatic Adjustments

The shill fees provide a consistent revenue stream, added to the high-risk portion to enhance trading capacity. The low-risk portion’s yield adds stability, while trading profits (if positive) can accelerate growth, with 50% reinvested.

  • If total treasury T < \$3,000 or high-risk portion H < \$1,600, reallocate to 70% low-risk, 30% high-risk to reduce exposure.
  • If T > \$5,000 and H>$2,000H > $2,000H > \$2,000, reallocate to 30% low-risk, 70% high-risk for growth.
  • For values between $3,000 and $5,000, maintain 50-50.

This dynamic reallocation ensures that as the treasury grows, it can take on more risk for higher returns, but if it shrinks, it reduces exposure to protect remaining capital. An unexpected detail is the integration of shill fees as a direct funding mechanism for high-risk trading, which is uncommon in traditional finance but aligns with Kaja’s community-driven model.

2. AI Decision-Making & Execution Strategy

  • How does Kaja evaluate which trades to execute? Does it rely purely on community sentiment, or does it incorporate market analysis (e.g., technical indicators, liquidity depth, volatility)?

    Kaja agent processes the following data to operate the treasury:

    • Market Data: Real-time prices, liquidity, and volatility from Conflux DEXs like Swappi and GraFun, and general market data from Dexscreener and CoinGecko (for bluechips).
    • Social Sentiment: Analysis of user shills and social media trends (from X) to evaluate meme coin potential, including engagements metrics within smart accounts (from Elfa and Yappers).
    • User Reputation: Scores based on past shill success to prioritize reliable suggestions, implemented through a reputation system to counter manipulation.
    • Risk Metrics: Volatility and liquidity thresholds to filter out risky trades, such as position sizing (max 5-10% of treasury per trade) and stop-loss mechanisms.
    • Yield Farming Data: Current APYs from Conflux DEXs and native staking.
  • Have any backtests been conducted on the AI’s trading strategy? If so, what success rates or key metrics have been observed?

    • Based on our current tests, high-risk portion performance correlates with the general meme coin market. Creators shills has a higher performance rate, cause it happens during the early stage of the coin gaining more potential upsides.
  • Given that the AI relies on external APIs (DEX tools, CoinGecko), how does the system handle service disruptions or inaccurate data feeds?

    • Kaja uses minimum two sources of data to secure itself from service disruptions (CoinGecko and Dexscreener in parallel). The same approach is used for RPCs and LLM APIs (using ChatGPT 4o and DeepSeek R1).

3. Security & Smart Contract Risks

  • Have the treasury management contracts undergone an independent security audit? If not, are there plans for an audit before launch or is an audit planned for later?
    • Kaja Treasury has no smart contracts, it operates with her threshold signature wallet. It has not passed the audit. Only internal security reviews has been conducted between several internal teams.
  • How is treasury custody managed? Since you mention an MPC wallet, what safeguards are in place to prevent unauthorized access or potential exploitation?
    • In order to secure the funds, we have a removed the wallet service from the agent core and built a system of permissions and rules on the wallet side to prevent misuse or potential attacks. So the basic treasury rules are parametrized on the wallet side.
  • Could Kaja be vulnerable to market manipulation via coordinated shilling attempts (e.g., a group of users artificially promoting illiquid assets)? How do you intend to mitigate this risk?
    • This risk is mitigated by market driven decisions. Kaja wont buy an asset if the market or the sentiment for it is bad, only if shiller will sponsor this trade as a donation:)
    • Of course some decisions may be a result of user’s manipulation, but in fact, Kaja uses community driven suggestions as a source of ideas, not as an advice. Playing with memes is a risky game and Kaja Treasury goal is to make it sustainable under different market conditions.

4. Growth & Engagement on Twitter/X

  • Kaja will be posting trades and engaging on Twitter—how will you ensure its tweets gain visibility beyond the existing Conflux community?
    • Based on our current research, meme coin creators are actively sharing any content (X posts) related to their coin, which brings a wider engagement (especially considering that GraFun has a multichain creators community). As an example: the latest user who shilled the meme (on March 31st) was https://x.com/CrumbCoinCTO - and this is already pure organic traffic.
  • What content strategy will Kaja follow? (E.g., meme engagement, alpha calls, AI-driven market insights, interactive polls?)
    • Kaja will follow a mixed strategy, blending AI-driven market insights (e.g., token buy/sell signals based on user suggestions) with interactive polls to engage users and gather input. Meme engagement could be incorporated to appeal to crypto culture, alongside occasional alpha calls to position Kaja as a valuable resource.
  • Are there any plans for collaborations with existing meme coin communities or influencers to drive visibility?
    • Kaja partners with launchpads (Grafun) in order to drive collaboration with large number of meme projects. Kaja would partner with trending projects and X influencers to amplify reach, leveraging its token-trading functionality as a unique hook.
  • How will you measure and optimize engagement metrics (e.g., likes, retweets, mentions) to track growth?
    • Engagement will be measured via chat participation on Kaja.app (DAU, MAU), likes, retweets, mentions on social media. Optimization could involve A/B testing content types (e.g., polls vs. insights), tracking peak activity times, and refining posts based on what drives the most interaction.
  • Given that AI-generated tweets often struggle with authenticity and relatability, how will Kaja’s Twitter activity be curated to ensure it sounds human-like and engaging rather than robotic?
    • To avoid sounding robotic, Kaja’s Twitter activity could be curated by blending AI-generated drafts with human oversight—tweaking tone, adding humor, or referencing current crypto trends. User suggestions from the chat would also be woven in to keep it conversational and relatable.
  • Are there any paid marketing campaigns planned for Twitter/X ads, or will growth be entirely organic?
    • We are planning to launch a KOLs campaign on Twitter / X (paid marketing) + social campaign (quiz) on Intract.xyz for the community to engage. Total campaign cost would be $1,000

5. Team & Long-Term Vision

  • While Zpoken appears to have expertise in cryptography and blockchain engineering, could you highlight specific team members with experience in AI-based trading systems?
  • Your proposal outlines a two-month roadmap. Could you provide additional details on the long-term plans post-launch? For example, how do you see Kaja evolving in the next 6–12 months? We do want the grants program to ba a great stepping stone for sustainable projects.
    • We extend the engagement campaigns with users and KOLs based on the Kaja Treasury performance and are considering to streamline fees generated from shills for this purpose.
    • Another direction for Kaja is to focus on low-risk portfolio portion, working within the ecosystem for more yield options and opportunities.