# Calculus Finance ## What is Calculus? Calculus is an automated, multi-chain liquidity management tool. In its initial release (V1), it supports BNB Smart Chain (BSC), helping users efficiently manage their LP positions with minimal effort and maximum control. Calculus Finance is **the first on-chain framework for training and deploying autonomous trading agents, built exclusively on BNB Chain (BSC) and backed by MH Ventures, CMS, Comma3 Ventures, Sidedoor Ventures, and TPC.** Using inverse reinforcement learning on the full decision histories of top-performing wallets and traders, we distill real on-chain behavior into agents that autonomously manage liquidity, TGE/memecoin trading, and spot/perp strategies within user-defined risk limits. Our first product line, **Agent α**, will launch first on BNB Chain, turning BNB-native LP positions into 24/7 self-optimizing market-making agents and deepening liquidity across key BNB DeFi venues. We’d love to explore a joint BNB × Calculus rollout and integrations to position BNB Chain as the home of the first large-scale on-chain trading-agent economy. X: Website: ## 1 Overview **Calculus Finance is the first fully on-chain framework for training autonomous trading agents.** Using inverse reinforcement learning, it extracts the underlying strategic patterns from the complete behavioral histories of top traders and major on-chain whale wallets. These extracted behavioral patterns are used to train specialized investment agents, which then evolve continuously through user preference alignment, self-play, and multi-agent reinforcement learning.
## 2 Product Architecture We consider the crypto market a highly complex real-time environment—one with strong adversaries, evolving rules, and pervasive noise. Building on this foundation, we developed a closed-loop evolutionary system—drawing inspiration from AlphaGo’s self-play paradigm—to train autonomous trading agents capable of producing commercial-grade alpha. Our training system is built on three pillars: #### (1) Alpha Seeker Dataset — Expert Behavior Cloning * Systematically capture and structure the full decision lifecycle of elite wallets and quantitative entities: position building, dynamic position management, cross-chain/cross-protocol arbitrage, liquidity manipulation, order-book dynamics, and more. This dataset provides “expert decision flows” far richer than public K-line data. * Users gain access to this master dataset, but customize it with their own trading and risk preferences. Agents operate autonomously within strict user-defined risk frameworks (max drawdown, leverage limits, position concentration). Users only receive periodic performance attribution and key decision explanations. #### (2) Agent Arena — Massively Parallel Self-Play Environment * Tens of thousands—even hundreds of thousands—of agent instances trade simultaneously in real market environments. * Agents share a standardized interface: observation space, action space, and reward function. * Evolution Strategies (ES), Population-Based Training (PBT), and adaptive compute allocation are used at scale. * Strong agents receive more compute and reproduction rights; weak agents mutate or are eliminated. * Evolution happens 24/7, continuously improving strategic capability. #### (3) On-Chain Darwinian Selection — Evolution Through Real Capital * Real capital drives the evolution of strategies, beginning with small test allocations and scaling progressively. * Profitability determines survival likelihood, compute allocation, and the transfer of evolved strategic parameters. * All decisions and execution paths are on-chain verifiable, fully auditable, inheritable, and forkable. By combining these components, Calculus transitions from traditional “improved user experience” to a fully autonomous AI-native investment economy powered by self-evolving on-chain agents. Our agent family consists of three major product lines: ### Agent α — On-Chain Liquidity Management (Mainnet 2024 Q4) * One-click deposit into target liquidity pools with automated single-asset or multi-asset rebalancing. * Multiple expert strategy templates validated through billions of Arena self-play simulations. * Continuous monitoring of funding rates, price deviation, pool depth, and MEV; triggers optimal rebalancing within seconds. * Fully customizable strategy framework with a dashboard showing impermanent loss, cumulative fees, and performance curves. {% content-ref url="/pages/NRHXTi1gel8bDvqBfKt7" %} [Key Features](/welcome-to-calculus/key-features.md) {% endcontent-ref %} {% content-ref url="/pages/nP3LpLvCdQjov2PFpSAH" %} [How It Works](/welcome-to-calculus/how-it-works.md) {% endcontent-ref %} {% content-ref url="/pages/trkEfrU6hqz5iSRJcihc" %} [Who Can Benefit?](/welcome-to-calculus/who-can-benefit.md) {% endcontent-ref %} {% content-ref url="/pages/hOhFb8CAioXJJFxn1dVQ" %} [Security and Audits](/welcome-to-calculus/security-and-audits.md) {% endcontent-ref %} {% content-ref url="/pages/Zn8eb9fxvYlAdwj4RyUh" %} [Fee Structure](/welcome-to-calculus/fee-structure.md) {% endcontent-ref %} {% content-ref url="/pages/x0fm5RhBXmJMJ1mExGD8" %} [Getting Started](/welcome-to-calculus/getting-started.md) {% endcontent-ref %} {% content-ref url="/pages/UKgSzCBZR5gq9e4Jl40M" %} [FAQ](/welcome-to-calculus/faq.md) {% endcontent-ref %} {% content-ref url="/pages/bh21zQJ7ECy4ykLxAvLL" %} [Calculus Seed User Campaign: DBTI Battle Royale](/welcome-to-calculus/calculus-seed-user-campaign-dbti-battle-royale.md) {% endcontent-ref %} {% content-ref url="/pages/0JkpD4jfG6xosyigy1ha" %} [Calculus Scoring System](/welcome-to-calculus/calculus-scoring-system.md) {% endcontent-ref %} ### Agent β — TGE & Memecoin Trading (2025 H1) * Pre-TGE positioning via DEX 1–4 hours ahead of launch. * Dynamic execution during TGE using order-book depth, funding rates, and sell-pressure distributions. * Introduces non-price factors for meme trading: buyer-cluster behavior, Telegram/Discord sentiment, KOL position flows. * Operates autonomously under user-defined risk parameters; users receive periodic performance reports only. ### Agent γ — Spot & Perpetual Trading (TBA) ## 3. Roadmap #### 2025 Q2–Q4 * Agent α mainnet launch * DBTI Arena * Alpha Seeker Dataset (v1) + limited Arena simulation access #### 2026 Q1–Q2 * Agent β mainnet launch * Agent Arena launch * Expanded set of non-price predictive features #### 2026 Q3–Q4 * Agent γ mainnet launch * Multi-Agent Coordination Protocol * Data incentive framework release #### 2027 * TEE + ZKP-powered verifiable compute * No-code agent creation * Agent-to-Agent paid invocation * Model-weight marketplace --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://calculusfinance.gitbook.io/welcome-to-calculus/calculus-finance.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.