> For the complete documentation index, see [llms.txt](https://docs.noelclaw.fun/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.noelclaw.fun/getting-started/getting-started.md). # Quickstart New to Noelclaw? You'll be up and running in under 5 minutes. *** ## What is Noelclaw? Noelclaw is an MCP skill - a plugin for AI tools like Claude, Cursor, Bankr, Aeon, and Hermes. Once installed, your AI gets 121 tools: persistent memory that carries across every session, autonomous agents that run on a schedule, live market data, DeFi execution on Base, deep research, GitHub integration, token scanning, code generation, audit trails, workflow packets, and more. You talk to it naturally. No commands to memorize. *** ## Step 1 - Check Node.js Noelclaw requires Node.js 18 or newer. Check if you have it: ```bash node --version ``` If you see `v18.x.x` or higher - you're good. If not, download it free from [nodejs.org](https://nodejs.org) (choose the LTS version). *** ## Step 2 - Install the Skill ### One-command setup (recommended) ```bash npx -y -p @noelclaw/mcp@3.43.1 noelclaw install ``` This detects all MCP-compatible apps on your machine (Claude Desktop, Cursor, Windsurf, VS Code, Zed) and configures each one automatically. No JSON editing. Then restart your client. *** ### Manual setup - pick your client ### Claude Desktop **Mac** - open this file in any text editor: ``` ~/Library/Application Support/Claude/claude_desktop_config.json ``` **Windows** - open this file: ``` %APPDATA%\Claude\claude_desktop_config.json ``` Paste this (or add the `noelclaw` block if the file already exists): ```json { "mcpServers": { "noelclaw": { "command": "npx", "args": ["-y", "-p", "@noelclaw/mcp@3.43.1", "noelclaw-mcp"] } } } ``` Save the file. **Fully quit and restart Claude Desktop** - not just close the window. → [Detailed Claude guide](/install-by-client/claude-install.md) *** ### Claude Code (terminal) ```bash claude mcp add noelclaw -s user -- npx -y -p @noelclaw/mcp@3.43.1 noelclaw-mcp ``` That's it. Restart Claude Code and the tools are available. *** ### Hermes ```bash hermes mcp add noelclaw -- npx -y -p @noelclaw/mcp@3.43.1 noelclaw-mcp ``` Or add it to `~/.hermes/config.yaml`: ```yaml mcp_servers: noelclaw: command: npx args: - -y - -p - "@noelclaw/mcp@3.43.1" - noelclaw-mcp timeout: 30 ``` Then type `/reload-mcp` in any Hermes session. → [Detailed Hermes guide](/install-by-client/hermes-openclaw.md) *** ### Cursor / Windsurf Edit `~/.cursor/mcp.json` (Cursor) or `~/.windsurf/mcp_config.json` (Windsurf): ```json { "mcpServers": { "noelclaw": { "command": "npx", "args": ["-y", "-p", "@noelclaw/mcp@3.43.1", "noelclaw-mcp"] } } } ``` → [Detailed Cursor guide](/install-by-client/cursor-install.md) *** ## Step 3 - Try It Out Once installed, just talk to your AI normally. No tool names needed - Noelclaw picks the right tool automatically. **Search the web in real time:** > "Find the latest news about AI agents today" **Check the market:** > "What's ETH doing right now?" **Ask for analysis:** > "Give me a full bull vs bear thesis on ETH" **Set up an autonomous monitor:** > "Monitor AI agents and automation news every morning at 8am" **Check your vault:** > "Show me my recent research" **Save something to memory:** > "Remember that I prefer low-risk DeFi and only trade on Base" **Test memory recall in a new chat:** > "What do you know about my trading preferences?" **Scan for opportunities:** > "Are there any tokens showing dip reversals right now?" **Run a simulation:** > "Simulate how markets would react if the Fed cuts rates 100bps" *** ## Step 4 - Autonomous Monitors (the best part) Noelclaw can run research automatically on a schedule - no prompting needed. ``` "Monitor AI agents and automation news every morning at 8am" ``` What happens next: * A scheduled job is registered at `daily-8am` * Every morning, Noelclaw searches the web for your topic * Summarizes the findings with an urgency score (1–5) * Saves the full report to your vault * Sends a Telegram briefing - quiet on routine days, loud on breaking news * Compares each run to the previous one, highlighting what changed To manage monitors: > "What monitors do I have running?" "Cancel my morning brief monitor" To get Telegram delivery, run `node worker/scripts/setup-telegram.mjs` once to connect your bot. *** ## Step 5 - Semantic Memory Noelclaw remembers things you tell it - across sessions, across chats. Tell it once: > "I prefer Lido for staking and Aerodrome for LP. I avoid leverage and meme coins." Come back in a new chat tomorrow and ask: > "What yield strategies make sense for me?" It already knows. No need to repeat yourself every session. → [Full memory guide](/memory-and-knowledge/memory.md) *** ## Step 6 - Noel Shell (Webapp Tool Calling) When you use Noelclaw in the webapp at [app.noelclaw.com](https://app.noelclaw.com), the chat supports **native tool calling** via Noel Shell. Instead of just answering questions, the chat can take actions: * **Spawn agents** — "Start an agent to monitor DeFi yields" * **Save to vault** — "Save this analysis to my vault" * **Search memory** — "What do I already know about ETH?" * **Create automations** — "Set up a DCA buying $50 of ETH every day" * **Estimate swaps** — "How much USDC do I need for 0.5 ETH?" * **List agents** — "Show me my active agents" * **Check wallet balance** — "What's my Base balance?" Seven agents are available in the webapp: **Noel** (crypto), **CoinGecko** (crypto data), **Sage** (analysis), **Forge** (developer), **Quill** (creative), **Spectre** (trading), and **Atlas** (general). → [Noel Shell docs](/agents-and-workflows/noel-shell.md) *** ## Troubleshooting | Problem | Fix | | ----------------------------------- | -------------------------------------------------------------------------- | | Tools not showing in Claude Desktop | Make sure you fully quit and restarted (not just closed the window) | | `npx: command not found` | Node.js isn't installed - download from [nodejs.org](https://nodejs.org) | | First message is slow | Normal - first run downloads the package (\~5 seconds). Fast after that. | | Tool call fails with auth error | Make sure you're signed in at [app.noelclaw.com](https://app.noelclaw.com) | | Vault tools return empty | Normal on first use - start by saving something with `vault_save` | | `schedule_research` not working | Add `TRIGGER_SECRET_KEY` to your MCP config env block | | Monitor runs but no Telegram | Add `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` to your MCP config env | --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## 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, and the optional `goal` query parameter: ``` GET https://docs.noelclaw.fun/getting-started/getting-started.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. 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.