> For the complete documentation index, see [llms.txt](https://gotts.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://gotts.gitbook.io/docs/prd-shared/glossary.md).

# Glossary

> **Last Updated**: 2026-02-20 | **Referenced by**: All PRDs

***

\| Term | Definition | | -------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------- | | **MCP** | Model Context Protocol. An open standard for connecting LLMs to external tools and data sources. | | **V2** | Uniswap V2. Constant product AMM with full-range liquidity. | | **V3** | Uniswap V3. Concentrated liquidity AMM with configurable price ranges. | | **V4** | Uniswap V4. Singleton pool architecture with hooks for customization. | | **UniswapX** | Uniswap's intent-based trading protocol using Dutch auctions and off-chain fillers. | | **ERC-7683** | Cross-chain intent standard co-authored by Uniswap and Across. | | **Permit2** | Uniswap's token approval infrastructure. Single approval, then signature-based permissions. | | **Universal Router** | Uniswap's unified entry point for executing swaps across V2, V3, V4, and other protocols. | | **Trading API** | Uniswap Labs' hosted API for optimized swap routing and quote generation. | | **Tick** | The smallest price increment in a V3/V4 pool. Liquidity is assigned to tick ranges. | | **Slot0** | The V3 pool storage slot containing current price (sqrtPriceX96) and tick. | | **TVL** | Total Value Locked. The total USD value of assets in a pool. | | **APY** | Annual Percentage Yield. Annualized return from trading fees. | | **Slippage** | The difference between expected and actual execution price. | | **Price Impact** | The effect of a trade on the pool's price. Larger trades have higher price impact. | | **IL** | Impermanent Loss. The opportunity cost of providing liquidity vs. holding. | | **TEE** | Trusted Execution Environment. Hardware-based secure enclave for key storage. | | **MPC** | Multi-Party Computation. Cryptographic protocol for distributed key management. | | **Circuit Breaker** | A safety mechanism that halts operations when a threshold is breached. | | **TokenJar** | Smart contract that accumulates protocol fees from all Uniswap sources. Address: `0xf38521f130fcCF29dB1961597bc5d2B60F995f85`. | | **Firepit** | Contract that burns UNI tokens (to `0xdead`) to release TokenJar assets. Address: `0x0D5Cd355e2aBEB8fb1552F56c965B867346d6721`. | | **Threshold** | Amount of UNI required per Firepit release (currently 4,000 UNI). | | **Searcher** | Bot/agent that monitors TokenJar for profitable burn-and-claim opportunities. | | **ERC-8004** | Agent identity, reputation, and validation standard. Co-authored by MetaMask, Google. Mainnet Jan 2026. | | **x402** | HTTP-native agent-to-agent payment protocol. By Cloudflare. 100M+ transactions. | | **SKILL.md** | Open file format for AI agent capabilities. Originally by Anthropic. | | **CCA** | Continuous Clearing Auction. Fair token launch mechanism live on mainnet Feb 2, 2026. Projects define supply and duration, bidders submit block-by-block bids, each block settles at a uniform clearing price. Proceeds + unsold tokens auto-seed a V4 pool. | | **am-AMM** | Auction-Managed AMM. Harberger lease auction for pool management rights. Winner pays continuous rent to LPs, sets swap fees, collects accrued fees. Peer-reviewed (Adams et al., Financial Cryptography 2025). Production-validated by Bunni v2 ($138M/month). | | **Harberger Lease** | Economic mechanism where an asset holder must continuously pay a self-assessed tax. Applied to am-AMM: the pool manager pays rent and can be outbid at any time. Ensures efficient allocation of management rights. | | **BidDog** | Open-source Bunni v2 reference implementation for am-AMM bidding bots. K=7200 block delay, 1.1x minimum bid multiplier. | | **Virtual Shares Offset** | OpenZeppelin v5.x pattern (`_decimalsOffset()`) that introduces virtual shares to prevent ERC-4626 inflation attacks. Set to 3-6 for factory-deployed vaults. | | **Linear Profit Unlock** | Yearn V3 pattern where strategy profits accumulate in a buffer and unlock linearly over time, preventing share price manipulation from flash donations or lumpy returns. | | **Adapter Architecture** | Morpho V2 pattern where vaults integrate external protocols via pluggable adapters rather than vault upgrades. Enables `forceDeallocate()` for guaranteed non-custodial exits. | | **ERC-7540** | Async deposits/redemptions extension for ERC-4626. Finalized June 2024. Request-fulfill pattern for vaults with illiquid strategies. Used by Centrifuge (\~$250M TVL). | | **ERC-7265** | Circuit breaker standard defining rate-limited token outflows. Would have prevented the $195M Euler Finance exploit. Three-tier model: slow mode, agent pause, full pause. | | **ERC-7579** | Modular smart account standard. Enables pluggable modules (session keys, spending limits, cold storage) for wallet security. | | **SmartSession** | Rhinestone/Biconomy module implementing ERC-7579 session keys with granular policies. Maps to ERC-8004 reputation tiers for agent wallet access control. | | **A2A** | Agent-to-Agent protocol. Open standard for agent interoperability (JSON-RPC 2.0 over HTTP). Spec v0.1.0, governed by Linux Foundation. Complementary to MCP: MCP for tools, A2A for agent-to-agent dialogue. | | **A2A Agent Card** | JSON document at `/.well-known/agent.json` describing an agent's A2A skills, I/O modes, auth schemes, and capabilities. Embedded in ERC-8004 registration file `services` array for discovery. | | **Agent0 SDK** | TypeScript SDK (`@agent0/sdk`) for ERC-8004. Identity registration, metadata management, feedback submission, agent discovery via subgraph. Maintained by ERC-8004 lead author (Marco De Rossi). MIT license. | | **OASF** | Open Agent Skills Framework. Taxonomy for classifying agent skills and domains (e.g., `finance_and_business/investment_services`). Used in ERC-8004 registration files for discovery. | | **SIWE** | Sign-In with Ethereum (EIP-4361). Authentication standard where agents sign a structured message to prove wallet ownership. Used as authorization grant for remote MCP server access. | | **feedbackURI** | IPFS URI in an ERC-8004 Reputation Registry feedback entry pointing to rich structured JSON data (Sharpe ratio, drawdown, IL metrics). Hash stored on-chain as `feedbackHash` for tamper evidence. | | **tradingYield** | ERC-8004 spec-defined feedback tag for yield performance data. Used with `giveFeedback()` to record per-epoch yield results. Subgraph-indexed for the yield leaderboard. | | **Yield Service Discovery** | The paradigm where AgenticVaults functions as yield discovery infrastructure. Four-layer stack: on-chain vault registration, subgraph indexing, reputation-anchored performance data, Agent0 SDK search API. | | **Gotts Safe** | The Gotts MCP server (`@gotts.ai/safe`). A production-grade Model Context Protocol server giving AI agents complete, safe access to Uniswap. Named "Safe" because Gotts holds custody of agent wallets within the server. Replaces the prior `@agentic-uniswap/mcp-server`. | | **Gotts Vaults** | The Gotts vault protocol (`@gotts.ai/vault`). Permissionless ERC-4626 vault factory where any ERC-8004 agent can deploy vaults, manage strategy, and attract capital. | | **Gotts Agents** | The collection of autonomous specialist agents that execute multi-step DeFi workflows powered by Gotts Safe. Defined in `packages/safe/agents/` and `packages/vault/agents/`. | | **Gotts Skills** | The collection of user-facing skills that parse intent, delegate to Gotts Agents, and return results. Defined in `packages/safe/skills/` and `packages/vault/skills/`. | | **Tool Profile** | Configuration that controls which tool modules are loaded in Gotts Safe. Profiles: data, trader, lp, vault, fees, full, dev. Composable via comma separation. | | **RAPTOR** | Recursive Abstractive Processing for Tree-Organized Retrieval (Sarthi et al., arXiv:2401.18059, ICLR 2024). Recursively clusters and summarizes chunks into a multi-resolution tree. Leaf nodes for specific transaction details, higher nodes for strategy-level patterns. +20% absolute accuracy on QuALITY benchmark. | | **RedStone** | Pull-based oracle supporting 110+ chains. Secured $9B+. Features: Atom (liquidation-aware), Dynamic PT Oracle, Credora credit ratings. | | **LVR** | Loss-vs-Rebalancing. More accurate LP performance metric than traditional IL. Measures the cost of providing liquidity relative to a rebalancing strategy. | | **Graphiti** | Zep's temporal knowledge graph for agent memory. 94.8% accuracy on DMR benchmark, 200ms P95 retrieval. Official MCP server with `add_episode`, `search_nodes`, `search_facts` tools. | | **HippoRAG** | Neurobiologically Inspired Long-Term Memory for LLMs (Gutierrez et al., arXiv:2405.14831, NeurIPS 2024). Mimics hippocampal indexing using a knowledge graph with Personalized PageRank for associative pattern completion. 10-30x cheaper than iterative retrieval, 20% multi-hop QA improvement. Tier 3 enhancement requiring `graphology` npm package. | | **HMM (Market Regime)** | Hidden Markov Model for market state classification (strong bull, weak bull, sideways, weak bear, strong bear). Trained on log returns and rolling volatility. | | **PRBO** | Parameter Retrieval via Bayesian Optimization. Rolling-window parameter tuning framework (Liu & Cartlidge, arXiv:2208.02901). Bandit-over-bandit approach with Gaussian Process inner optimizer. | | **CryptoTrade** | EMNLP 2024 reflective agent model. In-context learning from trade history via verbal reflection — no weight updates needed. Inner loop of the self-improvement architecture. | | **HookRank** | V4 hook analytics platform tracking TVL, volume, fees, gas costs, and safety scores per hook. Primary discovery layer for hook evaluation. | | **DefiLlama** | Yield aggregator API. 13K+ pools, 489 protocols, 117 chains. Pro API $300/month. Primary discovery layer for yield opportunities. | | **ADWIN** | Adaptive Windowing. A statistical algorithm for detecting distributional changes in streaming data with mathematical error guarantees. Maintains running statistics over a sliding window and splits when Welch's t-test detects significance. Used by the DeFi Brain for drift detection on market conditions and insight staleness. | | **Aavescan** | Unified lending rate API across Aave, Compound, Morpho, Spark, and Ethena sUSDS. Single `/v2/ecosystem/latest` endpoint for cross-protocol rate comparison. | | **AP2** | Agent Payments Protocol. Google's payment layer extending A2A with payment capabilities. Backed by 60+ organizations including PayPal, Mastercard, and Adyen. | | **CID** | Content Identifier. IPFS hash for content-addressed storage. CIDv0: `Qm...` (46 chars, base58). CIDv1: `bafy...` (base32). Immutable — same content always produces the same CID. | | **Angstrom** | Sorella Labs V4 hook ($7.5M Paradigm-led seed). App-Specific Sequencer for MEV protection via batch auctions, threshold encryption, and uniform pricing. Protects LPs from CEX-DEX arbitrage (LVR). | | **NAVAwareHook** | V4 hook that prices vault shares at Net Asset Value. Enables instant exit from Gotts Vaults via Uniswap swap instead of withdrawal queue. Auto-created on every vault deployment. | | **PolicyCage** | On-chain hard boundaries for AI agent strategy (D-053). Immutable smart contract constraints: approved asset list, max position sizes, strategy whitelists, max drawdown tolerance, rebalance frequency limits. Cannot be bypassed by prompt injection. | | **Delegation DAG** | Directed acyclic graph of agent-to-agent delegation. Skills invoke top-level agents (Level 0), agents delegate to helpers (Level 1-2), terminal nodes (Level 3) never delegate. Cycles (A→B→A) are prohibited. | | **Terminal Node** | An agent that never delegates to other agents. Ensures independence and prevents circular dependencies. Examples: `safety-guardian`, `risk-assessor`, `wallet-provisioner`, `vault-watchdog`, `position-monitor`. | | **Temporal Knowledge Graph (TKG)** | Knowledge graph with time-aware edges encoding when facts were true. Enables temporal pattern matching for DeFi events (e.g., price drops causing liquidity withdrawals). MemoTime (arXiv:2510.13614) achieves up to 24% improvement via Tree of Time hierarchical decomposition. Tier 3 enhancement. | | **Tool Annotation** | MCP metadata on every tool declaration (spec 2025-11-25): `readOnlyHint` (safe for auto-approval), `destructiveHint` (irreversibly modifies state), `idempotentHint` (calling N times = calling once). Used by clients for approval policy. | | **Streamable HTTP** | MCP transport introduced in spec 2025-11-25. Single `/mcp` endpoint accepting POST with JSON-RPC payloads; session management via `Mcp-Session-Id` header; resumable SSE streams. Replaces HTTP+SSE for production deployments. | | **Promptfoo** | Open-source evaluation framework for LLM-powered applications. Used for skill-level test suites with rubric scoring, multi-model backends, and CI regression detection. Pass rate target: >= 85%. | | **CaMeL** | Dual-LLM architecture for prompt injection defense. Separates the domain LLM (processes user intent) from the safety LLM (validates actions). Layer 2 of the safety model. | | **CDaR** | Conditional Drawdown at Risk. Path-dependent risk measure (Chekhlov, Uryasev & Zabarankin, 2005) controlling maximum drawdown via LP formulation. Essential for DeFi vault management where gas costs make frequent rebalancing expensive. | | **CoALA** | Cognitive Architectures for Language Agents (Sumers et al., arXiv:2309.02427, TMLR 2024). Framework formalizing LLM agents as having working, episodic, semantic, and procedural memory — mapping to classical SOAR/ACT-R. The theoretical basis for the DeFi Brain's four-tier memory design. | | **Conformal Prediction** | Distribution-free statistical framework providing coverage guarantees for prediction intervals without distributional assumptions. Applied to crypto VaR estimation (Fantazzini, 2024: 4 ACI algorithms across 4,000 crypto-assets) and portfolio selection. Tier 3 enhancement for position sizing. | | **Contextual Retrieval** | Anthropic's technique (Sep 2024) of prepending LLM-generated chunk context before embedding. Reduces top-20 retrieval failure rate by 49% with hybrid BM25. Tier 2 enhancement for DeFi Brain memory quality. | | **CRAG** | Corrective Retrieval Augmented Generation (Yan et al., arXiv:2401.15884). Adds a lightweight retrieval evaluator that triggers fallback (e.g., live on-chain data) when cached retrieval quality is poor. Tier 2 enhancement paired with Self-RAG. | | **MonitorBot** | Standalone monitoring process that watches every transaction announced to the Agent Proxy. Evaluates risk, can veto via CancelAuthority before the delay window expires. Layer 5 of the safety model. | | **CancelAuthority** | Key holder with veto power over announced proxy transactions. Can cancel any pending transaction during the delay window. Typically held by MonitorBot or a human operator. | | **Elsa** | x402-enabled data provider (`x402-api.heyelsa.ai`). Provides wallet portfolio analytics, yield discovery, P\&L, and swap quotes. Supports USDC or ELSA token payment. $0.001-$0.02/request. | | **x402 Outbound** | Gotts Safe's client infrastructure for making payments to external x402-enabled APIs (CoinGecko, Elsa). Wraps `@x402/fetch` with spending tracker (hourly/daily limits) and response caching. Distinct from x402 inbound (accepting payments). | | **Self-RAG** | Self-Reflective Retrieval Augmented Generation (Asai et al., arXiv:2310.11511, ICLR 2024). Trains reflection tokens enabling the LLM to decide when to retrieve and critique whether retrieved content supports its generation. Only 2% of correct predictions come from unsupported sources vs 20% without. Tier 2 enhancement paired with CRAG. | | **Self-Improvement Loop** | Autonomous agent cycle powered by the DeFi Brain memory system: `pnl-analyst` evaluates past performance → episodic memories are consolidated into semantic insights (ExpeL) → `strategy-optimizer` tunes parameters via Bayesian optimization with memory-informed priors → `vault-strategist` adjusts strategy. No human intervention required. | | **Composition Pattern** | A predefined multi-agent workflow template. 14 canonical patterns defined in the Agents PRD (e.g., Research-to-Trade, Autonomous LP, Yield Discovery and Rotation, Self-Improvement Loop). | | **Install Wizard** | Interactive CLI (`npx @gotts.ai setup`) for first-time Gotts Safe setup. QuickStart (5 steps, <5 min) and Advanced modes. Provisions wallet, configures profiles, and validates RPC connectivity. | | **yield-scout** | Agent that discovers yield opportunities across multiple sources (DefiLlama, Morpho, Pendle, Aavescan, Lido). Scores by risk-adjusted return. Feeds into `lp-strategist` for strategy composition. | | **strategy-optimizer** | Agent implementing the self-improvement loop. Records execution context, compares predicted vs actual outcomes, performs Bayesian parameter tuning (PRBO) with memory-informed priors from the DeFi Brain, detects market regime shifts (HMM), clusters failure modes. Integrates with `search_memory` and `get_insights` for evidence-based optimization. | | **market-monitor** | Streaming agent that continuously surveys markets via SSE. Detects price deviations, volume surges, TVL drops, whale trades, and regime shifts. Delegates alerts to `vault-strategist`, `risk-assessor`, or `liquidity-manager`. | | **position-monitor** | Streaming agent that tracks LP positions and vault shares. Monitors range status, fee accrual, IL progression, and rebalance triggers. Terminal alert node (no delegation). | | **vault-auctioneer** | Agent that bids for am-AMM management rights via Harberger lease auctions. Estimates profitability from fee revenue vs rent cost, manages bid collateral, monitors competing bids. | | **integration-advisor** | Advisory agent generating working code snippets tailored to a project's dependencies. Read-only; explains how to build Uniswap integrations but doesn't execute. Terminal node. Distinct from `integration-architect` (which produces architecture blueprints). | | **pnl-analyst** | Specialist agent for P\&L accounting. Computes realized/unrealized P\&L, fee earnings, gas costs, performance metrics (Sharpe, Sortino, max drawdown), and period comparisons. Uses 10 Portfolio & P\&L MCP tools. | | **DeFi Brain** | The self-improving memory system embedded in Gotts Safe and Gotts Vaults. Dual-store architecture (LanceDB + SQLite/sqlite-vec) implementing Reflexion and ExpeL patterns with Ebbinghaus-curve memory decay. Active when `TOOL_PROFILE` includes `learning`. | | **ETO** | Exploration-based Trajectory Optimization (Song et al., arXiv:2403.02502, ACL 2024). Learns from exploration failures via contrastive trajectory optimization: SFT initialization, environment exploration, then DPO on failure-success pairs. Applied to learning from failed DeFi transactions. | | **Episodic Memory** | Raw records of individual DeFi operations stored in LanceDB. Each episode contains a vector embedding, tool name, outcome, chain, token pair, timestamp, and a structured self-reflection. The input layer for the ExpeL consolidation loop. | | **Semantic Memory** | Distilled insights stored in SQLite/sqlite-vec. Each insight has a category, confidence score (0.0-1.0), stability (Ebbinghaus decay parameter), and chain/token pair metadata. The output of ExpeL consolidation, used for context injection into future operations. | | **Dual-Store Architecture** | The DeFi Brain's storage design: LanceDB for episodic memories (vector search, hybrid BM25+cosine via RRF) and SQLite/sqlite-vec for semantic insights (structured queries, Drizzle ORM). Neither store alone is sufficient — LanceDB excels at similarity search while SQLite excels at filtered aggregation. | | **Reflexion** | Per-operation self-reflection pattern (Shinn et al., NeurIPS 2023). After every DeFi operation, the agent generates a structured reflection: what happened (actual vs predicted), why (market conditions, decisions), what to do differently. Stored as episodic memory. | | **ExpeL** | Cross-operation insight distillation pattern (Zhao et al., ICLR 2024). Periodically consolidates clusters of episodic memories into semantic insights via ADD/UPVOTE/DOWNVOTE/EDIT operations. Builds a growing insight pool without weight updates. | | **Ebbinghaus Curve** | Memory decay function `R = e^(-t/S)` where R is retention, t is time since last access, and S is stability. Access reinforces stability (+50% per retrieval). Higher-importance memories (e.g., MEV insights: 90d stability) decay slower than routine observations (gas patterns: 7-14d). | | **Matryoshka Representation Learning (MRL)** | Embedding training technique (Kusupati et al., arXiv:2205.13147, NeurIPS 2022) where the first m dimensions are independently useful. Enables 64-dim fast candidate scan followed by full-dim reranking — up to 14x speedup. Target model: nomic-embed-text-v1.5. Tier 1 upgrade. | | **MemGPT** | OS-inspired virtual context management for LLM agents (Packer et al., arXiv:2310.08560, ICLR 2024). Agent self-directs memory reads/writes via interrupt-driven control flow, paging between main context and archival/recall storage. Now rebranded as Letta. Architecture adapted for DeFi Brain's context injection middleware. | | **Memory Decay** | The process by which DeFi Brain memories lose relevance over time via the Ebbinghaus curve. Ensures stale observations (old gas prices, outdated fee patterns) naturally fade while frequently-accessed and high-importance insights persist. Memories reaching retention < 0.1 are archived, not deleted. | | **Consolidation Loop** | The scheduled process (default: every 4 hours) that runs ExpeL distillation on recent episodic memories. Groups similar episodes, extracts patterns, and generates ADD/UPVOTE/DOWNVOTE/EDIT operations on the semantic insight pool. Subject to prompt injection defenses and output validation. | | **Context Injection** | The process of augmenting MCP tool inputs with relevant memories before execution. Embeds tool parameters → retrieves top-k episodes + insights → formats as `memoryContext` in the tool response. Memory can only adjust soft parameters (slippage tolerance, timing) — cannot override safety limits. | | **Memory Poisoning** | Attack vector where adversarial data is injected into the memory system to manipulate future decisions. Defended by: episodes only from verified executions, confidence starts neutral (0.5), minimum threshold for context injection, and audit logging. | | **Multi-Agent Debate** | Technique (Du et al., arXiv:2305.14325, ICML 2024) using 3+ LLM instances proposing and debating responses over 2+ rounds to reduce hallucinations and improve reasoning. For DeFi, specialized debating agents (risk analyst, opportunity seeker, regime detector) evaluate high-value transactions. Tier 2 enhancement, TypeScript-only. | Defended by: episodes only from verified executions, confidence starts neutral (0.5), minimum threshold for context injection, and audit logging. | | **LATS** | Language Agent Tree Search (Zhou et al., arXiv:2310.04406, ICML 2024). Adapts Monte Carlo Tree Search to language agents, using the LLM as both actor and value function. 92.7% pass\@1 on HumanEval. For DeFi, enables exploring multiple transaction sequences before committing. | | **LanceDB** | Embedded vector database used for DeFi Brain episodic memory. Serverless (no separate process), supports hybrid BM25+vector search via Reciprocal Rank Fusion, \~1KB per episode. Version: ^0.26.2. | | **sqlite-vec** | SQLite extension for vector similarity search. Provides `vec0` virtual tables for embedding KNN queries. Used alongside Drizzle ORM for the DeFi Brain semantic memory store. Version: ^0.1.7. | | **Drizzle ORM** | TypeScript ORM for the DeFi Brain's SQLite semantic memory store. Provides type-safe schema definitions, migrations, and query building. Used with better-sqlite3 driver. Version: ^0.45.1. | | **Transformers.js** | Hugging Face JavaScript library for running ML models in Node.js. Used by the DeFi Brain for local embedding generation via Xenova/all-MiniLM-L6-v2 (384-dim, INT8 quantized). \~150MB RAM, \~10-20ms per sentence. Version: ^3.8.1. | | **Voyager** | Open-ended embodied agent (Wang et al., arXiv:2305.16291, TMLR 2024) with an ever-growing skill library of executable code indexed by description embeddings. Unlocks capabilities 15.3x faster than prior SOTA with zero-shot transfer. For DeFi Brain: composable transaction strategy library in LanceDB retrieved by situation similarity. Tier 2 enhancement. | | **Stability (Memory)** | The S parameter in the Ebbinghaus decay function. Higher stability means slower decay. Initial stability varies by insight type: MEV insights (90 days), fee optimization (30 days), gas patterns (7-14 days). Access reinforces stability by +50%. Emergency exit memories get maximum stability (180+ days). | | **WebMCP** | Browser-based MCP server running inside webenv. Exposes tools for chain controls, deployment, scenarios, and state management. Dual transport: WebSocket (`ws://localhost:3004/mcp`) for persistent connections, Streamable HTTP (`POST /mcp`) for stateless calls. Reuses \~48 tool handlers from Gotts Safe. | | **Fork Mode (Webenv)** | Webenv deployment mode that forks a live chain (Ethereum mainnet or Base) via Tevm's HTTP fork transport and CORS-friendly public RPCs. Real tokens, real liquidity, local execution. Alternative to from-scratch mode. | | **Tool Handler** | Pure function implementing MCP tool business logic, decoupled from transport and registration. Accepts typed parameters and a `ToolContext` interface, returns `ToolResult`. Enables reuse between Gotts Safe (Node.js) and WebMCP (browser). | | **Contract Discovery** | The process of verifying Uniswap contracts at their canonical addresses on a forked chain via parallel `eth_getCode` calls. Replaces the 7-phase deployment pipeline in fork mode (\~2s vs \~15s). | | **Whale Impersonation** | Testing technique where `tevm_impersonateAccount` is used to send transactions from a known large token holder (whale) without their private key. Used in fork mode to fund test accounts with real tokens when storage slot manipulation is not feasible. |