SuperAPI AI Directory

Arsenal-Quant-Project

Deterministic MEV and slippage risk oracle for DeFAI agents.

Open website

About this tool

Arsenal-Quant-Project (Arsenal Decision Engine) is a risk verification layer for autonomous AI agents (DeFAI), focused on providing deterministic MEV, Impermanent Loss, and slippage risk metrics for DeFi operations.

Purpose: Before an autonomous agent deploys capital or adjusts a standard constant product/full range LP position (such as Uniswap V2 or full range V3), it submits pool parameters (APY, price ratio, holding days) to the API, and the engine calculates the exact mathematical risk, net return (R_net), and dynamic Breakeven Corridor boundaries. The engine does not run a stateful trading bot, nor does it generate speculative prediction signals; instead, it provides a stateless risk middleware layer for agents to query before acting.

Main capabilities:

  • Deterministic calculation: The breakeven corridor is a deterministic algebraic boundary (IL = cumulative return). Any position whose price ratio remains within [lower_be, upper_be] has R_net > 0 by mathematical definition, not by a probabilistic model.
  • No LLM, no hallucination: Pure algebraic calculation, O(1) time and memory complexity, local execution latency below 15ms.
  • The engine measures rather than predicts: It does not claim prediction accuracy, and the methodology and raw results are published (backtest script, 180 days of Binance ETH/USDC daily close data).
  • Response JSON contract: The layer field reports how the service was called, FREE means within the free quota, PREMIUM means L402 payment has been verified.
  • Each response carries an HMAC tag over the result, used to detect tampering with the engine's own output; this is a symmetric provenance marker, not proof that a third party can independently verify.

Usage: Two invocation paths are provided:

  1. MCP JSON-RPC: Standard MCP handshake initialize → tools/list → tools/call, supporting streamable HTTP and stdio.
  2. REST convenience routes: No MCP client required. Both routes run the same calculation and the same quota. Note that /mcp/evaluate only supports GET; POST to that path returns 405 Allow: GET, because JSON-RPC belongs to /mcp. Developer integration: The integration manual and MCP guide are in COOKBOOK.md; the MCP auto-discovery card is at https://api.arsenal-quant.com/.well-known/mcp/server-card.json。提供 Python integration example.

Applicable scenarios: Suitable for autonomous DeFi agents to perform risk verification before deploying or maintaining standard constant product/full range LP positions, helping agents deterministically determine whether a position is above or below the breakeven boundary.

Limitations and current status:

  • Free tier: evaluate_pool is 100 calls per IP per day, including custom parameters, no Lightning wallet required; this is currently the only tier in service.
  • Paid tier: The gateway implements the L402 challenge, returning 402 and a WWW-Authenticate header when the free quota is exceeded. However, the payment channel is not yet running in production; invoices issued today cannot be settled, and payment is neither expected nor accepted. The paid tier should be regarded as announced but unavailable.
  • GET /mcp/audit/latest: 3 free calls per IP per hour, returns 402 when exceeded; that response only documents the protocol and is not a live payment path.
  • The engine does not prevent losses, nor does it claim how much money it can save.
  • The intended model is per-call pricing, so that cost can be budgeted like any other input, but this model is not yet in service, and all calls are currently free.