> 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/gotts-safe-mcp-server/mcp-server/13-distribution.md). # Distribution and Synergies > **Package**: `packages/safe/` (`@gotts.ai/safe`) | **Prerequisites**: [01-overview.md](/docs/gotts-safe-mcp-server/mcp-server/01-overview.md) > > For supported chains reference, see [shared/chains.md](/docs/prd-shared/chains.md). For glossary, see [shared/glossary.md](/docs/prd-shared/glossary.md). *** ## Synergies ### Companion Gotts Skills (Separate PRDs) Gotts Safe is the execution layer. Gotts Skills provide workflow guidance and best practices. These are authored as separate PRDs (see [Gotts Skills PRD](/docs/skills/skills.md)) and live in `packages/safe/skills/` (or `packages/vault/skills/` for vault skills). | Skill | Plugin | Status | Relationship to Gotts Safe | | -------------------- | ----------------------- | ------ | ----------------------------------------------------------------------------------------------------------- | | `manage-liquidity` | `packages/safe/skills/` | Spec | LP guidance and execution skill. Agents use Gotts Safe LP tools for execution. | | `setup-agent-wallet` | `packages/safe/skills/` | Spec | Wallet setup guidance for agents. Covers all 7 wallet types Gotts Safe supports. | | `build-hook` | `packages/safe/skills/` | Spec | V4 hook development guidance. Separate from Gotts Safe (hooks are smart contracts, not runtime operations). | | `manage-treasury` | `packages/safe/skills/` | Spec | Agent treasury management. Uses Gotts Safe swap and LP tools for fee conversion and yield. | | `deploy-agent-token` | `packages/safe/skills/` | Spec | Agent token deployment. Uses Gotts Safe pool creation and LP tools. | | `agent-market-maker` | `packages/safe/skills/` | Spec | Agent market making on own token pairs. Uses Gotts Safe LP tools. | ### Companion Gotts Agents (Separate PRDs) | Agent | Purpose | Status | Gotts Safe Usage | | ---------------------- | --------------------------------------------------------- | ------ | ----------------------------------------------------------------------------------------------------------------------------------------- | | `lp-strategist` | Recommends LP ranges and strategies | Spec | Calls `get_pool_info`, `get_tick_data` for analysis | | `portfolio-analyst` | Watches positions, alerts when out of range | Spec | Calls `get_position`, `get_positions_by_owner` | | `treasury-manager` | Autonomous treasury management for self-funding agents | Spec | Calls `execute_swap`, `add_liquidity`, `collect_fees` for fee conversion and yield | | `token-deployer` | Automated token pool creation and liquidity bootstrapping | Spec | Calls `add_liquidity`, `get_pool_info` for V4 pool creation | | `identity-verifier` | ERC-8004 agent identity verification | Spec | On-chain reads to ERC-8004 registries | | `agent-service-broker` | Agent-to-agent service matching and settlement | Spec | Calls `execute_swap` for payment settlement | | `protocol-fee-seeker` | Autonomous TokenJar monitor and burn executor | Spec | Calls `get_tokenjar_balances`, `get_firepit_state`, `execute_burn`, `subscribe_tokenjar`, `get_burn_history`, `get_fee_accumulation_rate` | | `trade-executor` | Handles full swap lifecycle with retries | Spec | Calls `get_quote`, `execute_swap`, `check_safety_status` | ### Related Products (Separate PRDs) | Product | Relationship | | ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------ | | Agent Safety Toolkit (library) | Extracted from MCP server's safety middleware as a standalone npm package. Other agent builders can use safety primitives without the full MCP server. | | Intent-Based Agent Trading SDK | High-level SDK that wraps Gotts Safe's tools into a TypeScript API (`agent.swap()`, `agent.addLiquidity()`). | | Autonomous LP Manager | Background service that uses Gotts Safe's LP tools to manage positions autonomously. | ### Vault Integration (Unified via Profiles) The Gotts Vaults (`packages/vault/`) is the primary value-multiplier for Gotts Safe. With the **profile system**, vault tools register directly into the core server when `TOOL_PROFILE` includes `vault` -- no separate server process needed. See [vault synergies](/docs/gotts-vaults/vault/16-synergies.md) for the full analysis. | Integration | How It Works | | ---------------------------- | -------------------------------------------------------------------------------------------------------- | | **Profile-based activation** | `TOOL_PROFILE=vault` registers all vault tools inside the core server. One deployment, one health check. | | **Shared safety pipeline** | Vault tools use the core server's full 15-layer safety pipeline directly (no optional import needed) | | **Pool data delegation** | `vault-strategist` agent delegates to `pool-researcher` and `risk-assessor` via core MCP tools | | **LP composition** | Vault rebalancing uses `add_liquidity`, `remove_liquidity`, `rebalance_position` from core tools | | **Share pool trading** | Vault share tokens trade on V4 pools via `execute_swap` | | **Volume flywheel** | Vault creation auto-deploys V4 pool -> active management -> LP volume -> share trading -> more volume | The standalone vault server (`packages/vault/src/server.ts`) remains available as a fallback for teams requiring complete isolation, but the recommended deployment is the unified server with the `vault` profile. ### Agent Capital Markets Integrations Gotts Safe should integrate with the emerging Agent Capital Markets stack: | Integration | Purpose | Priority | | -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- | | **x402 payment layer** | Agents pay per MCP request in USDC (\~$0.001-0.01 per call). Eliminates API key onboarding. Revenue from every agent interaction. | High (G11) | | **ERC-8004 identity** | Verified agents get lower fees, priority execution, access to premium pools. Agent swap history feeds into reputation scores. | High (G12) | | **Agent registry listing** | Register Gotts Safe as a discoverable service on agent registries. Agent discovery without manual configuration. | Medium | | **Clanker/BankrBot integration** | Automated pool creation when agent tokens are deployed. 585K+ tokens deployed through Clanker, all routing through Uniswap V4. | Medium | | **Agent treasury tools** | Tools for self-funding agents to auto-convert fees, DCA, and manage operating capital (G13). Serves BankrBot's 220K+ active wallets. | Medium | | **Vault protocol** | ERC-4626 agent-gated vaults for collective LP management, am-AMM bidding, and CCA participation. See [Vault PRDs](https://github.com/wpank/gotts.ai-monorepo/blob/main/prd/mcp-server/vault/README.md). | High (G20) | | **AP2 protocol** | Google's Agent Payment Protocol for payment authorization. 60+ partners including Visa, Mastercard. Future consideration. | Low | ### x402 Pricing Model Tiered pricing for MCP tool access via x402 micropayments (USDC on Base, \~200ms settlement): | Tier | Price | Tools | | ---------------- | -------------- | ------------------------------------------------------------------------------ | | **Free** | $0 | All read-only tools: `get_pool_info`, `get_token_price`, `search_tokens`, etc. | | **Micropayment** | $0.001 - $0.01 | Execution tools: `execute_swap`, `add_liquidity`, `submit_cca_bid`, etc. | | **Premium** | $0.01 - $0.10 | Strategy recommendations, historical analytics, am-AMM bid optimization | | **Subscription** | $10 - $100/mo | Streaming access, priority execution, vault strategy API | Reference implementation: Civic's `@civic/x402-mcp` pattern for Streamable HTTP + per-tool pricing maps via `createPaidMcpHandler()`. ### x402 Outbound Client Infrastructure Gotts Safe has x402 **inbound** infrastructure (accepting payments from other agents) and x402 **outbound** infrastructure (making payments to external x402-enabled APIs). The outbound client supports multiple providers (CoinGecko, Elsa) through a generic client pattern, enabling agents to acquire external data by paying per-request in USDC on Base. **Architecture:** ``` MCP Tool (e.g., get_onchain_token_data, get_wallet_portfolio) │ ▼ x402 Outbound Client (src/x402/client.ts) │ - Wraps @x402/fetch for automatic 402 payment flow │ - Spending tracker enforces hourly/daily limits (shared across providers) │ - Response caching prevents duplicate payments │ ├───────────────────────────────┐ │ │ ▼ ▼ CoinGecko x402 API Elsa x402 API (src/providers/coingecko.ts) (src/providers/elsa.ts) │ Token data, trending, │ Portfolio, wallet analytics, │ pool search │ yield, P&L, swap quotes │ $0.01/req │ $0.001-$0.02/req │ │ Supports USDC or ELSA token ▼ ▼ Tool returns enriched data to agent ``` **Components:** | File | Purpose | | ------------------------------ | ---------------------------------------------------------------------- | | `src/x402/client.ts` | Generic x402 client wrapping `@x402/fetch` for any x402 API | | `src/x402/types.ts` | TypeScript types for x402 configuration, spending, and providers | | `src/x402/spending-tracker.ts` | Rolling-window spending tracker with hourly and daily limits | | `src/providers/coingecko.ts` | CoinGecko x402 provider -- token data, trending pools, pool search | | `src/providers/elsa.ts` | Elsa x402 provider -- portfolio, wallet analytics, yield, P\&L, quotes | **Spending Limits:** | Limit | Default | Config Key | | ------------------- | ------- | ------------------------------- | | Per-hour maximum | $1.00 | `X402_OUTBOUND_MAX_PER_HOUR` | | Per-day maximum | $10.00 | `X402_OUTBOUND_MAX_PER_DAY` | | Per-request maximum | $0.10 | `X402_OUTBOUND_MAX_PER_REQUEST` | **Configuration:** ```bash # Enable x402 outbound payments X402_OUTBOUND_ENABLED=true X402_OUTBOUND_PRIVATE_KEY=0x... # Wallet for USDC payments on Base # Spending limits (optional, defaults shown) X402_OUTBOUND_MAX_PER_HOUR=1.00 X402_OUTBOUND_MAX_PER_DAY=10.00 X402_OUTBOUND_MAX_PER_REQUEST=0.10 ``` Config file schema (add to `x402` section): ```json { "x402": { "inbound": { "...existing inbound config..." }, "outbound": { "enabled": true, "privateKey": "${X402_OUTBOUND_PRIVATE_KEY}", "maxSpendPerHour": 1.00, "maxSpendPerDay": 10.00, "maxSpendPerRequest": 0.10, "providers": { "coingecko": { "baseUrl": "https://api.coingecko.com", "enabled": true }, "elsa": { "baseUrl": "https://x402-api.heyelsa.ai", "enabled": true, "paymentToken": "usdc" } } } } } ``` **Dynamic Import:** The `@x402/fetch` and `@x402/evm` SDKs are imported dynamically so Gotts Safe starts without them when x402 outbound is disabled. When disabled, tools that depend on x402 (e.g., `get_onchain_token_data`, `get_wallet_portfolio`) return a clear `X402_DISABLED` error with instructions to enable. Each provider is independently toggleable via the `providers` config. **Graceful Degradation:** If x402 payment fails (insufficient USDC, network issue, spending limit hit), tools fall back to existing on-chain data sources where possible. The response includes a `degraded: true` flag and `fallbackReason` field. CoinGecko tools fall back to on-chain RPC reads. Elsa tools fall back to Uniswap-only data (e.g., `get_yield_opportunities` falls back to Uniswap pool APY data; `get_wallet_portfolio` falls back to multi-chain balance reads). Wallet analysis and P\&L have no on-chain fallback and return a clear error. **Payment Token Note:** Elsa supports payment in USDC (default, via `/api/` path) or ELSA token (via `/api/elsa/` path). Both paths return identical data. Configure via `paymentToken` in the Elsa provider config. CoinGecko only supports USDC. ### Integration with Existing Toolkit Plugins Gotts Safe will be registered in the toolkit's plugin system: * **Plugin**: A new plugin (e.g., `uniswap-onchain`) or integration into the existing `uniswap-integrations` plugin * **MCP config**: Added to the plugin's `.mcp.json` for auto-discovery by Claude Code * **Skills that invoke the server**: Companion skills use the MCP tools via `allowed-tools` in their frontmatter ### MCP Config Integration The server can be invoked two ways in `.mcp.json`. The **direct** approach (`@gotts.ai/safe`) is recommended for MCP client configs because it has faster cold-start (single package, no Commander routing overhead). The **unified CLI** approach (`@gotts.ai`) provides subcommand access to all Gotts tools (setup, safe, devenv, portal, etc.) through a single package. ```json { "mcpServers": { "uniswap": { "type": "stdio", // Direct (recommended for MCP configs — faster cold-start): "command": "npx", "args": ["-y", "@gotts.ai/safe@latest"], // Via unified CLI (same server, slightly slower startup): // "args": ["-y", "@gotts.ai@latest", "safe"], "env": { "WALLET_TYPE": "local", "PRIVATE_KEY": "${UNISWAP_AGENT_PRIVATE_KEY}", "TOKEN_ALLOWLIST_MODE": "strict", "SPENDING_LIMIT_PER_TX_USD": "10000", "SPENDING_LIMIT_DAILY_USD": "100000" } } } } ``` > **Note**: JSON does not support comments. The `//` lines above are for illustration only — use one `args` array or the other in actual configs. After global install (`pnpm add -g @gotts.ai`), the unified CLI is available as `gotts`: ```bash gotts safe # Start MCP server (stdio) gotts safe --http # Start MCP server (HTTP) gotts setup # Interactive install wizard (replaces npx @gotts.ai/create) gotts portal # Agent management dashboard (replaces npx @gotts.ai/portal) gotts devenv # Start local Uniswap testnet gotts doctor # Diagnose configuration issues ``` *** ## Competitive Comparison | Feature | This Server | uniswap-trader-mcp | uniswap-poolspy-mcp | uniswap-price-mcp | uniswap-pools-mcp | GOAT Uniswap Plugin | | --------------------------- | ----------- | ------------------ | ------------------- | ----------------- | ----------------- | ------------------- | | **Tools** | 84 (target) | 2 | 1 | 2 | 2 | \~3 | | **Chains** | 11 | 8 | 9 | 4 | Multiple | EVM | | **Protocol versions** | V2, V3, V4 | V3 only | V2, V3 | V3 only | V2, V3, V4 | V3 | | **LP management** | Full | No | No | No | No | No | | **UniswapX** | Yes | No | No | No | No | No | | **ERC-7683 cross-chain** | Yes | No | No | No | No | No | | **Permit2** | Yes | No | No | No | No | No | | **Pre-flight simulation** | Yes | No | No | N/A | N/A | No | | **Token allowlist** | Yes | No | No | N/A | N/A | No | | **Spending limits** | Yes | No | N/A | N/A | N/A | No | | **Rate limiting** | Yes | No | N/A | N/A | N/A | No | | **Balance circuit breaker** | Yes | No | N/A | N/A | N/A | No | | **Nonce management** | Yes | No | N/A | N/A | N/A | No | | **Wallet options** | 7 types | Raw key only | N/A | N/A | N/A | GOAT wallets | | **Uses Trading API** | Yes | No | No | No | No | No | | **Uses official SDKs** | Yes | No | No | No | No | Partial | | **Author** | Uniswap | kukapay | kukapay | kukapay | kukapay | Crossmint | ## Glossary > See [shared/glossary.md](/docs/prd-shared/glossary.md) for the full glossary of terms used across all PRDs. *** ## Distribution: Dual Delivery Model ### Why Dual Delivery OpenClaw (188K+ GitHub stars, 5,705+ community skills on ClawHub) is the dominant agent skills ecosystem but has **no native MCP client support** -- Issue #4834 (Jan 30, 2026) was closed as "not planned." Community bridges exist (`openclaw-mcp` with OAuth 2.1, `openclaw-mcp-plugin` with Streamable HTTP) but native support is absent. To reach the maximum number of agent frameworks, the Gotts Safe must be delivered in two formats: 1. **Native MCP server** (`@gotts.ai/safe` on npm) -- primary delivery for Claude Code, Cursor, Vercel AI SDK, and any MCP-compatible framework 2. **AgentSkills package** (`packages/safe/skills/` with SKILL.md) -- for OpenClaw/ClawHub, GitHub Copilot, and any AgentSkills-compatible framework ### AgentSkills Wrapping Strategy Each MCP tool category maps to an AgentSkills skill folder: | MCP Tool Category | AgentSkills Skill | Skill File | | -------------------- | ----------------------- | --------------------------------------------- | | Data and Analytics | `uniswap-pool-analysis` | `packages/safe/skills/uniswap-pool-analysis/` | | Trading | `execute-swap` | `packages/safe/skills/execute-swap/` | | Liquidity | `manage-liquidity` | `packages/safe/skills/manage-liquidity/` | | CCA and Token Launch | `deploy-agent-token` | `packages/safe/skills/deploy-agent-token/` | | Protocol Fees | `seek-protocol-fees` | `packages/safe/skills/seek-protocol-fees/` | Skills load progressively (name/description at startup, full instructions on demand) following the AgentSkills specification. ### Security Considerations Given the **ClawHub malicious skills incident** (283-386 malicious entries, ClawHavoc campaign delivering Atomic Stealer malware targeting crypto wallet keys): * All published AgentSkills packages must be signed and verified * VirusTotal scanning on every release * No embedded download instructions in SKILL.md (the "pastebin piping" attack vector) * Wallet key isolation in TEEs -- skills never have direct key access *** ## Source Repository Reference This PRD was authored in the `uni-ai` monorepo. ### Repository Structure ``` gotts-monorepo/ ├── packages/ │ ├── cli/ # @gotts.ai — Unified CLI (setup, safe, devenv, portal, etc.) │ │ └── src/ │ │ ├── cli.ts # Commander entry point, subcommand registration │ │ ├── commands/ # One file per subcommand (setup.ts, safe.ts, devenv.ts, etc.) │ │ └── utils/ # Shared CLI utilities │ ├── safe/ # @gotts.ai/safe — Gotts Safe MCP server (standalone) │ │ └── src/ │ │ ├── index.ts # Server entry point, tool registration │ │ ├── tools/ # One file per MCP tool │ │ ├── providers/ # Chain and subgraph providers │ │ ├── constants/ # Addresses, ABIs, subgraph URLs │ │ ├── config.ts # Configuration management │ │ ├── cache.ts # Response caching │ │ ├── errors.ts # Structured error types │ │ └── types.ts # Shared types │ ├── vault/ # @gotts.ai/vault — Gotts Vaults (standalone MCP + SDK + contracts) │ ├── tui/ # @gotts.ai/tui — Shared terminal UI primitives │ ├── shared/ # @gotts.ai/shared (types, errors, chain utils) │ └── agent-proxy/ # @gotts.ai/agent-proxy (time-delayed execution) ├── .claude/ # Claude Code config │ ├── commands/ # Slash commands (68) │ └── settings.json ├── prd/ # Product requirements documents │ ├── mcp-server/ # Gotts Safe PRD (this directory) │ ├── agents/ # Agents PRD │ ├── skills/ # Skills PRD │ ├── shared/ # Shared references (market context, glossary, chains) │ └── vault/ # Vault PRDs (16 files) ├── research/ # Market research and ecosystem analysis └── scripts/ ``` ### Skill File Format ```yaml --- description: Trigger phrases for auto-activation allowed-tools: Read, Write, Edit, Task(subagent_type:agent-name) model: opus --- # Skill Title [Markdown instructions, patterns, code examples] ``` ### Agent File Format ```yaml --- description: Agent expertise description model: opus allowed-tools: Read, Glob, Grep, WebFetch, WebSearch --- # Agent Role You are an expert in [domain]. ## Expertise Areas ``` ### plugin.json Format ```json { "name": "plugin-name", "version": "1.0.0", "skills": ["./skills/skill-name"], "commands": ["./skills/skill-name/skill-name.md"], "agents": ["./agents/agent-name.md"] } ``` ### MCP Server Config (.mcp.json) ```json { "mcpServers": { "uniswap": { "type": "stdio", "command": "npx", "args": ["-y", "@gotts.ai/safe"], "env": { "WALLET_TYPE": "local", "PRIVATE_KEY": "${KEY}" } } } } ``` > **Direct vs CLI**: Use `@gotts.ai/safe` (above) in MCP client configs for faster cold-start. Use `@gotts.ai` with subcommands (`npx @gotts.ai safe`, `npx @gotts.ai setup`, etc.) for interactive CLI usage. See [MCP Config Integration](#mcp-config-integration) for details. ### Naming Conventions | Component | Format | Pattern | Examples | | --------- | ----------- | --------------- | --------------------------------------- | | Skills | verb-noun | Action-oriented | `execute-swap`, `seek-protocol-fees` | | Agents | noun-role | Entity-oriented | `trade-executor`, `protocol-fee-seeker` | | MCP Tools | snake\_case | Resource-action | `get_token_price`, `execute_burn` | ### Uniswap SDK Packages `@uniswap/sdk-core`, `@uniswap/v3-sdk`, `@uniswap/v4-sdk`, `@uniswap/universal-router-sdk`, `@uniswap/router-sdk`, `@uniswap/permit2-sdk`, `@uniswap/uniswapx-sdk`, `@uniswap/smart-order-router` ### Reference Patterns for Local Testnet & DCA The local testnet and DCA tools draw on established patterns: * **Local testnet with mock Uniswap**: Anvil forks with mock V2 routers, test tokens, and whale impersonation for funding test accounts * **Time-travel testing**: `evm_increaseTime` + `evm_mine` for testing DCA cadences, LP fee accumulation, and governance timelocks without waiting * **V2/V3 adapter patterns**: Wrapping V3 pools behind V2-compatible interfaces for backward compatibility * **Permit2 integration**: Universal approval pattern for agents * **Environment management**: Single manifest → multi-target env generation (frontend, contracts, shell) * **DCA order lifecycle**: Create → execute → pause → resume → cancel with keeper automation via Gelato * **On-chain preview helpers**: `previewSwap()`, `previewOrderExecution()` for agent simulation before execution ### Protocol Fee System (Unification Proposal) The TokenJar/Firepit tools serve the Uniswap Unification proposal: * **TokenJar**: Central vault accumulating protocol fees from V2, V3, V4, UniswapX, and Unichain native fees * **Firepit**: Burns UNI tokens (to `0xdead`) in exchange for releasing TokenJar assets * **Fee source breakdown**: Each fee is attributed to its origin protocol * **Burn economics**: Threshold (currently 4,000 UNI), nonce for replay protection, profitability calculation * **Searcher patterns**: Monitor accumulated fees → calculate profit vs. burn cost → execute burn-and-claim *** ## npm Package Distribution ### Package Metadata Requirements Two npm packages serve different purposes: | Package | Purpose | Path | `bin` | | ---------------- | ---------------------------------------------------------- | --------------- | ------------------------------ | | `@gotts.ai/safe` | Standalone MCP server (recommended for MCP client configs) | `packages/safe` | `gotts-safe → ./dist/index.js` | | `@gotts.ai` | Unified CLI — Commander subcommands for all Gotts tools | `packages/cli` | `gotts → ./dist/cli.js` | #### `@gotts.ai/safe` (MCP Server) For discoverability on npm and validation by the official MCP Registry, `package.json` must include: ```json { "name": "@gotts.ai/safe", "mcpName": "io.github.gotts-protocol/safe", "bin": { "gotts-safe": "./dist/index.js" }, "keywords": [ "mcp", "model-context-protocol", "uniswap", "defi", "ai-agent", "claude", "cursor", "llm", "erc-4626", "base", "ethereum" ], "files": [ "dist", "README.md", "INSTALL.md", "CHANGELOG.md", "LICENSE", ".env.example" ] } ``` The `mcpName` field in `io.github.{org}/{repo}` format is **required** for the official MCP Registry. Without it, `mcp-publisher publish` validation fails and the server cannot be listed. #### `@gotts.ai` (Unified CLI) The unified CLI wraps all Gotts packages under a single `gotts` command. Subcommands dynamically import their target package to keep cold-start fast for each subcommand. ```json { "name": "@gotts.ai", "description": "Gotts — Agent Capital Markets powered by Uniswap. Unified CLI for setup, MCP server, devenv, portal, and more.", "bin": { "gotts": "./dist/cli.js" }, "keywords": [ "gotts", "uniswap", "defi", "ai-agent", "mcp", "model-context-protocol", "cli", "agent-economy", "erc-4626", "erc-8004" ], "files": ["dist", "README.md", "LICENSE"], "dependencies": { "commander": "^13.0.0", "@gotts.ai/core": "workspace:*", "@gotts.ai/tui": "workspace:*" } } ``` **Subcommands**: | Subcommand | Dynamic Import | Description | | ------------------- | ------------------------- | -------------------------------------- | | `gotts setup` | `@gotts.ai/safe` (wizard) | Interactive install wizard | | `gotts safe` | `@gotts.ai/safe` | Start MCP server (stdio or HTTP) | | `gotts devenv` | `@gotts.ai/devenv` | Start local Uniswap testnet | | `gotts portal` | `@gotts.ai/portal` | Agent management dashboard | | `gotts testnet` | `@gotts.ai/testnet` | Generic EVM test toolkit | | `gotts install-mcp` | (built-in) | Write MCP config to client config file | | `gotts config` | (built-in) | Manage Gotts configuration | | `gotts doctor` | (built-in) | Diagnose configuration issues | Heavy dependencies (`@gotts.ai/safe`, `@gotts.ai/devenv`, `@gotts.ai/portal`, `@gotts.ai/testnet`) are **dynamic imports** — they are only loaded when the subcommand is invoked, not at CLI startup. This keeps `gotts --help` instant. Bundled dependencies (`@gotts.ai/core`, `@gotts.ai/tui`) are lightweight and always available. **Deprecation mapping**: | Old Command | New Command | | ----------------------------- | ---------------------------- | | `npx @gotts.ai/create` | `npx @gotts.ai setup` | | `npx @gotts.ai/create --zero` | `npx @gotts.ai setup --zero` | | `npx @gotts.ai/create --ui` | `npx @gotts.ai setup --ui` | | `npx @gotts.ai/portal` | `npx @gotts.ai portal` | ### npm Distribution Tags Maintain separate npm distribution tags: | Tag | Published When | Config Example | | --------- | -------------------------- | ----------------------------------------- | | `@latest` | Stable releases | `"args": ["-y", "@gotts.ai/safe@latest"]` | | `@beta` | Preview features | For adventurous early adopters | | `@dev` | Latest main branch (CI/CD) | Internal testing only | All example configs in documentation use `@latest`. Advanced users can pin: `@gotts.ai/safe@1.0.0`. ### MCP-Specific SemVer Semantics Because MCP tool schemas are public contracts consumed by LLMs and agent frameworks, standard semver is applied with MCP-specific nuances: | Change | Version Bump | Reason | | ---------------------------------------- | ------------ | ----------------------------------------- | | Remove or rename a tool | **MAJOR** | Agents that depend on the tool will break | | Change a required parameter name or type | **MAJOR** | Tool invocations will fail | | Add a new tool | MINOR | Additive, backward compatible | | Add optional parameter with default | MINOR | Backward compatible | | Fix a tool bug without schema change | PATCH | No contract change | | Fix tool description | PATCH | No schema change | The `serverInfo.version` in the MCP `initialize` response **must always match `package.json`.** Clients cache tool lists by server name+version. ### `CHANGELOG.md` Format Maintain a `CHANGELOG.md` following [Keep a Changelog](https://keepachangelog.com) format. Every entry must call out tool-level changes explicitly: ```markdown ## [1.1.0] — 2026-03-01 ### Added - `get_trending_pools` — Top pools by 24h volume across all chains (data profile) ### Fixed - `get_token_price` handles tokens with no liquidity gracefully ## [1.0.0] — 2026-02-18 ### Added - Initial release: 27 tools across data, trader, lp, vault, fees profiles ``` ### Registry Listings For comprehensive ecosystem listings (official MCP Registry, Smithery, Docker MCP Catalog, mcp.so, PulseMCP, and one-click install deeplinks), see [**16-registries.md**](/docs/gotts-safe-mcp-server/mcp-server/16-registries.md). That document contains: * The full `server.json` for the official MCP Registry * The `smithery.yaml` configuration * The `server.yaml` for Docker MCP Catalog submission * Cursor and VS Code deeplink install button implementations * Complete npm package requirements *** ### Companion Research Documents | Document | Focus | | -------------------------------------------------- | ----------------------------------------------- | | `uniswap-agent-ecosystem-ideas.md` | Master research, prioritization, open questions | | `agent-to-agent-economy-2026.md` | x402, self-funding agents, value accrual | | `defi-agent-infrastructure-feb-2026.md` | Payment protocols, identity, coordination | | `ai-agent-security-safety-trust-2026.md` | Threats, safety patterns, wallet security | | `agent-hackathons-and-competitions-2026.md` | Hackathons, SKILL.md, launchpads | | `agent-economy-opportunities.md` | Full skills/agents/MCP roadmap, gap analysis | | `erc-8004-defi-agents-research.md` | ERC-8004 spec, V4 hooks, vaults | | `privy-agentic-wallets-infrastructure-feb-2026.md` | Privy architecture, policies | | `onchain-ai-agent-ecosystem-feb-2026.md` | Builder profiles, standards |