Agentic Primitives

Agentic Primitives are stateless, typed analytics functions that answer questions about DeFi positions, pools, and risk. They follow a three-line interface: stateless construction, computation at .apply(), typed dataclass return. Every primitive works identically whether called from a notebook or from an LLM via the MCP server. See The Primitive Interface for the interface invariants.

These primitives read pool state — they are analytics operations (position decomposition, risk assessment, price scenarios). For execution operations that mutate pool state (mint, burn, swap, zap), see Core Primitives.

Availability

Primitive	V2	V3	Balancer	Stableswap	MCP
`AnalyzePosition`	✅	✅	❌	❌	✅
`AnalyzeBalancerPosition`	❌	❌	✅	❌	✅
`AnalyzeStableswapPosition`	❌	❌	❌	✅	✅
`SimulatePriceMove`	✅	✅	❌	❌	✅
`SimulateBalancerPriceMove`	❌	❌	✅	❌	✅
`SimulateStableswapPriceMove`	❌	❌	❌	✅	✅
`CheckPoolHealth`	✅	✅	❌	❌	✅
`DetectRugSignals`	✅	✅	❌	❌	✅
`DetectFeeAnomaly`	✅	❌	❌	❌	❌
`CheckTickRangeStatus`	❌	✅	❌	❌	❌
`AssessDepegRisk`	❌	❌	❌	✅	✅
`OptimalDepositSplit`	✅	❌	❌	❌	❌
`EvaluateRebalance`	✅	❌	❌	❌	❌
`EvaluateTickRanges`	❌	✅	❌	❌	❌
`CompareFeeTiers`	❌	✅	❌	❌	❌
`CompareProtocols`	✅	✅	✅	✅	❌
`CalculateSlippage`	✅	✅	❌	❌	✅
`DetectMEV`	✅	✅	❌	❌	❌
`AggregatePortfolio`	✅	✅	✅	⚠️	❌
`FindBreakEvenPrice`	✅	✅	❌	❌	❌
`FindBreakEvenTime`	✅	✅	❌	❌	❌

Primitive descriptions

Each primitive links to the category page where it’s documented in detail with runnable examples and result-dataclass fields. Per-class API pages are in progress.

AnalyzePosition() — Decompose a V2/V3 LP position into impermanent loss, fee income, net PnL, and diagnosis.
AnalyzeBalancerPosition() — Position decomposition for Balancer weighted pools, accounting for asymmetric token weights.
AnalyzeStableswapPosition() — Position decomposition for Stableswap pools, with alpha/unreachable-alpha regime handling.
SimulatePriceMove() — Project a V2/V3 position’s value at a hypothetical price change.
SimulateBalancerPriceMove() — Project a Balancer position’s value at a hypothetical price change, respecting pool weights.
SimulateStableswapPriceMove() — Project a Stableswap position’s value at a hypothetical price change, respecting amplification.
CheckPoolHealth() — Snapshot pool-level health metrics: TVL, reserves, LP concentration, fee activity, spot price.
DetectRugSignals() — Threshold-based rug-pull signal detector. Composes over CheckPoolHealth.
DetectFeeAnomaly() — Detect unusual fee patterns in V2 pools.
CheckTickRangeStatus() — V3-only: is the position’s tick range still active?
AssessDepegRisk() — Stableswap-only: quantify exposure to a stablecoin depeg using the ε ↔ δ relationship.
OptimalDepositSplit() — Compute the optimal token split for a single-sided V2 deposit.
EvaluateRebalance() — Evaluate whether rebalancing a V2 position is worth the cost.
EvaluateTickRanges() — V3-only: compare candidate tick ranges for a position.
CompareFeeTiers() — V3-only: compare fee tier options for a position.
CompareProtocols() — Cross-protocol comparison of the same token pair across V2, V3, Balancer, and Stableswap.
CalculateSlippage() — Compute slippage and price impact for a proposed swap on a V2/V3 pool.
DetectMEV() — Detect potential MEV exposure for a swap on a V2/V3 pool.
AggregatePortfolio() — Aggregate N LP positions into a single portfolio-level view. Dispatches to per-protocol analyzers.
FindBreakEvenPrice() — Find the price at which IL equals fee income for a V2/V3 position.
FindBreakEvenTime() — Find the holding period at which fee income compensates IL for a V2/V3 position.

Relationship to other sections

Core Primitives and Agentic Primitives are both abstract — they dispatch across protocols without the caller knowing protocol details. The distinction is what they do:

Core Primitives mutate pool state (execution: mint, burn, swap, zap)
Agentic Primitives read pool state and return analytics (position analysis, risk, scenarios)

For the protocol-specific exchange classes underneath both layers, see Protocol API → Uniswap V2 and siblings.

Categories

Each category has its own page with executable examples and result-dataclass details: