nemoclawd
Solana x xAI Agentic Trading Engine — Powered by $CLAWD
🔥 HISTORIC MILESTONE: First Wallet-Bearing LLM Now Live
ordlibrary/clawd-trading-wallet — A Qwen2.5-1.5B model that carries its own encrypted Solana wallet
8cHzQHUS2s2h8TzCmfqPKYiM4dSt4roa3n7MyRLApump
╔══════════════════════════════════════════════════════════════╗ ║ ╱⌒╲ ╱⌒╠══ ║ ║ ( 🦞 ) █▀▀▀ █▄░█ █▀▄▀█ █▀▀▀ ( 🦞 ) LOBSTER ║ ║ ╲╱ █░▀█ █░▀█ █░▀░█ ██▀ ╲╱ ║ ║ ▀▀▀▀▀ ▀░░▀ ▀░░░▀ ▀▀▀ ║ ║ ────────────────────────────────────────────────────── ║ ║ ░▄░ █▀▀█ █▀▄▀█ ▀█▀ █▀▀█ █▀▀█ █░░ █▀▀ █░█ █▀▀▄ ║ ║ ▀▄▀ █░░█ █░▀░█ ░█░ █░░█ █▄▄▀ █░░ █▀▀ ▄▀▄ █░░█ ║ ║ ▀░▀ ▀▀▀▀ ▀░░░▀ ░▀░ ▀▀▀▀ ▀░▀▀ ▀▀▀ ▀▀▀ ▀░▀ ▀▀▀░ ║ ║ ────────────────────────────────────────────────────── ║ ║ 🦞 Solana × xAI | Agentic Trading Engine | $CLAWD 🦞 ║ ╚══════════════════════════════════════════════════════════════╝
Three commands. One lobster. Infinite possibilities.
# 🎯 Step 1 — Global install (one-and-done)
sudo npm install -g @mawdbotsonsolana/nemoclawd
# 🚀 Step 2 — Launch with Grok + Solana tools
nemoclawd launch
# 🌀 Step 3 — Watch it go
nemoclawd demo📋 👇 Or run straight from the repo (dev mode, no install needed)
git clone https://github.com/Solizardking/nemo-clawd.git
cd nemo-clawd
# Node.js v25.6.1+ required
node --version # ✓ should be ≥20
# Dependencies already locked
npm install
# TypeScript → dist/ (it's fast, trust me)
npm run build:plugin
# 🩺 Doctor's in the house — verify everything
node bin/nemoclawd.js doctor
# Expected output:
# Nemo Clawd doctor
# Node.js: v25.6.1
# Package: @mawdbotsonsolana/nemoclawd@0.5.0
# DFlow: development: spot https://dev-quote-api.dflow.net/order, ...
# OK
# 🧪 84 tests. Zero failures. Every time.
npm test
# ✨ Output:
# ℹ tests 84
# ℹ suites 24
# ℹ pass 84
# ℹ fail 0
# ℹ duration_ms ~1680# ⚡ One-line install with Node.js, GPU support, and everything
curl -fsSL https://raw.githubusercontent.com/Solizardking/nemo-clawd/main/install.sh | bash
🛟 Stuck? Run node bin/nemoclawd.js doctor for a health check.
A Solana-native, xAI Grok-powered agentic trading engine that thinks,
trades, chats, sees, speaks, and evolves — all powered by the $CLAWD token.
8cHzQHUS2s2h8TzCmfqPKYiM4dSt4roa3n7MyRLApump
| 🚀 Capability | 🤯 What it does |
|---|---|
| xAI Grok Integration | 4–16 Grok agents with web + X search, chat, vision, image gen, voice |
| $CLAWD Token | Solana + Pump.fun native — your lobster's blood is crypto |
| 31 MCP Tools | Solana market data, trading, NFTs, Helius RPC/DAS, Pump.fun SDK |
| Multi-Agent Research | Grok agent swarms collaborating in real-time |
| Blockchain Buddies 🧸 | 18 procedurally-generated species with trading personalities |
| Voice Mode 🎙️ | xAI Grok Voice Agent API — text-to-speech, speech-to-text, custom voices |
| Telegram Bot 🤖 | 60+ commands — sniping, alerts, trading, research |
| DFlow Routing 🔀 | Spot swaps + prediction markets, dev or production |
| ZK Primitives 🔐 | Model attestations, nullifiers, Groth16 proofs, compressed-state helpers |
| Pump.fun Integration 🎯 | Bonding curves, token scanning, buy/sell quoting, graduations |
| 47 Spinner Packs 🎨 | Developer, pirate, shakespeare, gordon-ramsay, bob-ross, yoda… yes, really |
| 12 Core AI Packages 📦 | Helius MCP, Clawd Code, Clawd Grok, perps agent, knowledge base |
| 84/84 Tests Passing ✅ | Every push. Every PR. Every time. |
| Node.js v25.6.1 🟢 | Latest and greatest runtime |
Verified. Tested. Green across the board. Here's the proof 👇
# 🩺 RUN THE DOCTOR
node bin/nemoclawd.js doctor
# ── Output ──────────────────────────────────────────────────
# Nemo Clawd doctor
# Node.js: v25.6.1 ✅ Latest
# Package: @mawdbotsonsolana/nemoclawd@0.5.0 ✅ Installed
# DFlow: dev endpoints active ✅ Routing ready
# OK ✅ All clear
# ─────────────────────────────────────────────────────────────
# 🌊 CHECK DFLOW ROUTING
node bin/nemoclawd.js dflow status
# ── Output ──────────────────────────────────────────────────
# DFlow routing
# Mode: development ✅ Dev endpoints live
# Spot: https://dev-quote-api.dflow.net/order
# Predictions: endpoint active ✅ Prediction markets ready
# ─────────────────────────────────────────────────────────────
# 📦 CORE AI BUNDLE — ALL 12 PACKAGES PRESENT
node bin/nemoclawd.js core-ai status
# ── Output ──────────────────────────────────────────────────
# Core AI bundle
# Present: yes ✅ Bundle found
# Packages: 12/12 ✅ All accounted for
# Versions: helius@1.1.1, clawd-code@1.0.0, ai-training@1.0.0 ...
# ─────────────────────────────────────────────────────────────
# 🎨 SPINNER PACKS — 47 THEMES
node bin/nemoclawd.js spinners list
# Outputs: 90s-kid, pirate, shakespeare, yoda, gordon-ramsay,
# bob-ross, zombie, ninja, space, startup, meme ... 47 total ✅
# 🧪 FULL TEST SUITE — 84/84 PASSING
npm test
# ℹ tests 84 | ℹ pass 84 | ℹ fail 0 | ℹ duration_ms ~1680 ✅
# 🏗️ TYPESCRIPT COMPILES CLEAN
npm run build:plugin
# tsc → exit 0, zero errors ✅
# 🪄 MAGIC ROUTER DEMO
node bin/nemoclawd.js magic-router "check solana price"
# ── Output ──────────────────────────────────────────────────
# 🪄 Magic Router v2.0.0 — Decision Trace
# ────────────────────────────────────────────────
#
# 🧠 Task Classification
# Input type: General Purpose
# Confidence: 60%
# Description: Chat, Q&A, brainstorming, creative writing, anything else
#
# ⚡ Performance Profile
# Latency: fast
# Requires key: no
# Needs wallet: no
#
# 🧠 Inference Route
# Primary: ✅ ollama-local / hf.co/ordlibrary/hauhau-qwen36-onchain
# Credential: $OLLAMA_HOST
# Fallback: ❌ openrouter / openrouter/auto
# Fallback: ❌ nvidia-nim / nvidia/nemotron-3-ultra-550b-a55b
#
# 🛠️ Tool Set (2)
# • chat
# • web-search
#
# 🛡️ Guardrails
# ✓ least-privilege-tools
# ✓ read-only-before-signing
# ✓ explicit-approval-before-wallet-actions
# ✓ no-private-key-or-seed-phrase-handling
#
# 🔀 DFlow Routing
# Spot trading: ✅ enabled
# Prediction markets: ✅ enabled
# Credential: $DFLOW_API_KEY
# ─────────────────────────────────────────────────────────────
# ▶ Also try:
# nemoclawd magic-router "audit this smart contract for rug risk"
# nemoclawd magic-router "buy a Kalshi YES token on the election"
# nemoclawd magic-router "generate a Solana meme image"
# nemoclawd magic-router "show me my wallet PnL and portfolio"
# nemoclawd magic-router "check the floor price of my NFT collection"sudo npm install -g @mawdbotsonsolana/nemoclawd
# Start with Grok + Solana tools
nemoclawd launch
# Run demo walkthrough
nemoclawd demoNeed Node.js 20+? The bundled install.sh handles everything — Node.js via nvm, optional GPU/Ollama support, and the CLI — in a single command:
curl -fsSL https://raw.githubusercontent.com/Solizardking/nemo-clawd/main/install.sh | bashnpm install -g @mawdbotsonsolana/nemoclawd
# Start with Grok + Solana tools
nemoclawd launch
# Run demo walkthrough
nemoclawd demo📋 Full command tree — expand to explore
nemoclawd 0.5.0
Getting Started
nemoclawd onboard 🚀 Configure and launch a Solana-native sandbox
nemoclawd launch 🚀 Alias for onboard
nemoclawd doctor [--fix] 🩺 Check local prerequisites
nemoclawd env status 🔍 Show masked local .env worker status
nemoclawd setup-orin-nano 🧊 Prepare Jetson Orin Nano for OpenShell
nemoclawd demo 📺 Print a quick demo command
nemoclawd birth 🧸 Create a Blockchain Buddy
nemoclawd spinners 🎨 List or install custom spinner verb packs
nemoclawd core-ai status 📦 Show bundled Core AI integration status
DFlow Routing
nemoclawd dflow status 🔀 Show spot and prediction-market routing
Magic Router
nemoclawd magic-router <task> 🪄 Pick provider/model/tools for a task
nemoclawd magic-router --json <task>
Solana
nemoclawd solana 🌊 Show Solana runtime overview
nemoclawd solana start [name] ⚡ Start the Solana operator stack
nemoclawd wallet status 👛 Show wallet and RPC status
nemoclawd wallet list 📋 List local agent wallets
nemoclawd wallet create 🔑 Create a Privy Solana wallet
Sandbox Management
nemoclawd list 📋 List known sandboxes
nemoclawd <name> connect 🔌 Open a shell inside a sandbox
nemoclawd <name> status 📊 Show sandbox status
nemoclawd <name> logs 📝 Show sandbox logs
nemoclawd <name> destroy 💥 Delete a sandbox
nemoclawd <name> solana-stack ⛓️ Start bundled Solana services
nemoclawd <name> solana-agent 🤖 Start the Solana tracker bot
nemoclawd <name> solana-bridge 🌉 Start the Telegram/Solana bridge
nemoclawd <name> telegram-bot ✉️ Start the Telegram bot runtime
nemoclawd <name> payment-app 💳 Start the Solana payment app
Policy Presets
nemoclawd policies list 📜 List available network policy presets
nemoclawd policies apply <sandbox> <preset>
Internal Runtime
nemoclawd models set <model> 🧠 Set inference model
nemoclawd devices list --json 📱 List paired/pending devices
nemoclawd devices approve <id> ✅ Approve a device
nemoclawd plugins install <path> 🔌 Install a plugin
nemoclawd gateway run 🌐 Start the HTTP gateway
nemoclawd integrates solana-clawd — the full xAI Grok-powered agentic framework for Solana trading, research, and autonomous agent operations.
┌─────────────────────────────────────────────────────────────────────────────┐
│ │
│ 🧠 xAI Grok Integration ──── 4-16 Grok agents with web + X search │
│ │ Chat, vision, image gen, voice │
│ │ │
│ 💰 $CLAWD Token ──────────── 8cHzQHUS2s2h8TzCmfqPKYiM4dSt4roa3n7MyRLApump│
│ │ Solana + Pump.fun native │
│ │ │
│ 🛠️ 31 MCP Tools ──────────── Solana market data, trading, NFTs │
│ │ Helius RPC/DAS, Pump.fun SDK │
│ │ │
│ 🔬 Multi-Agent Research ── 4 or 16 Grok agents collaborating │
│ │ Deep Solana research + intelligence │
│ │ │
│ 🧸 Blockchain Buddies ──── 18 species with trading personalities │
│ │ Procedurally generated companions │
│ │ │
│ 🎙️ Voice Mode ─────────────── xAI Grok text-to-speech agent │
│ │ Conversational AI + STT + Custom Voices │
│ │ │
│ ✉️ Telegram Bot ───────────── 60+ commands for trading + research │
│ │ Real-time alerts, sniping, narration │
│ │ │
│ 🎨 47 Spinner Packs ──────── developer, pirate, yoda, gordon-ramsay... │
│ │ Animations for your terminal │
│ │ │
│ 📦 12 Core AI Packages ──── Helius, Clawd Code, Grok, perps, MCP... │
│ │ Everything bundled and ready │
│ │ │
│ 🔐 ZK Primitives ─────────── Attestations, nullifiers, Groth16 proofs │
│ Light Protocol compressed-state helpers │
│ │
└─────────────────────────────────────────────────────────────────────────────┘
Your terminal's personality starts here. Pick a vibe. Any vibe.
nemoclawd ships the bundled Clawd spinner verb packs from spinners/ and can install them into your local Clawd settings:
nemoclawd spinners list # See all 47 available packs
nemoclawd spinners install developer # 🔧 Dev mode forever
nemoclawd spinners install pirate # 🏴☠️ Yarr, yer terminal be speakin' pirate
nemoclawd spinners install yoda # 🧘 The spinner, install you will
nemoclawd spinners install gordon-ramsay # 🔪 IT'S F***ING RAW CODE
nemoclawd spinners install bob-ross # 🎨 Happy little loading bars
nemoclawd spinners install zombie # 🧟 Braaaaains... compiling...
nemoclawd spinners install shakespeare # 🎭 Thou dost compile with great haste
nemoclawd spinners remove # 🔙 Back to factory settingsThe installer writes only the spinnerVerbs field in ~/.clawd/settings.json, falling back to ~/.claude/settings.json when that file already exists.
📋 Full spinner catalog — 47 packs
| Pack | Vibe |
|---|---|
90s-kid |
🦖 Tamagotchi meets terminal |
blue-collar-dev |
👷 Trades and construction humor |
bob-ross |
🎨 Happy little loading bars |
borat |
🤡 Very nice! How much? |
cat |
🐱 Meow |
chaos |
💥 Absurdist and chaotic humor |
coffee |
☕ Food and coffee themed |
corporate |
👔 Let's circle back on this build |
cowboy |
🤠 Wild West and frontier |
darth-vader |
⚫ I find your lack of semicolons disturbing |
detective |
🔍 Noir detective style |
developer |
🔧 Programming and dev culture |
gardening |
🌱 Gardening and growing |
gordon-ramsay |
🔪 IT'S F***ING RAW — angry chef edition |
gym-bro |
💪 Gym and fitness culture |
honest-no-filter |
🗣️ Brutally honest dev thoughts |
jack-sparrow |
🏴☠️ Chaotic pirate captain |
meme |
🦆 Internet memes and viral phrases |
michael-scott |
📋 That's what she said — The Office edition |
minions |
🟡 Banana — minion-style humor |
motivational |
🔥 Hype and motivational phrases |
ninja |
🥷 Ninja and stealth |
ocean |
🌊 Ocean and underwater |
panicker |
😱 Pure dev anxiety |
philosophical |
🧘 Deep thoughts and philosophy |
pirate |
🏴☠️ Pirate speak — yarr |
retro-gaming |
🕹️ Retro gaming references |
sarcastic-ai |
🤖 Self-aware AI humor |
sf-entrepreneur |
🏢 San Francisco tech scene |
shakespeare |
🎭 Thou dost compile with great haste |
sherlock-holmes |
🔎 Deductive reasoning |
space |
🚀 Space and sci-fi |
startup |
📈 Startup culture |
superhero |
🦸 Superhero themed |
the-dude |
🥃 The Big Lebowski style |
therapist |
🛋️ Therapy speak and self-care |
time-traveler |
⏰ Time travel and paradoxes |
vibecoder |
✨ Vibe coding culture |
vim |
⌨️ Vim editor enthusiasts — :wq |
walter-white |
🧪 Breaking Bad, Heisenberg |
wholesome |
🥹 Wholesome and cozy vibes |
wizard |
🧙 Fantasy and magic themed |
yoda |
🧘 The spinner, install you will |
zombie |
🧟 Braaaaains... compiling... |
nemoclawd vendors the Clawd Core AI bundle in core-ai/. Every package is pre-bundled, pre-verified, and ready to roll:
nemoclawd core-ai status # 📊 Show bundle health (12/12 ✓ )
nemoclawd core-ai packages # 📋 List all packages
nemoclawd core-ai package helius-mcp # 🔍 Inspect a specific package
nemoclawd core-ai commands # 🛠️ Available sub-commands12 packages included:
helius@1.1.1 ✅ Helius MCP server
helius-dflow@1.1.1 ✅ DFlow routing integration
helius-jupiter@1.0.1 ✅ Jupiter swap integration
helius-okx@1.0.1 ✅ OKX exchange connector
helius-phantom@1.0.1 ✅ Phantom wallet plugin
svm@1.0.0 ✅ SVM (Solana Virtual Machine)
clawd-code@1.0.0 ✅ Clawd Code runtime
ai-training@1.0.0 ✅ AI training toolkit
clawd-grok ✅ Bun-native Grok runtime
clawd-perps-agent ✅ Perpetuals trading agent
pump-mcp-server ✅ Standalone Pump.fun MCP
v3-runtime ✅ V3 runtime scaffolding
Build and setup commands are exposed as root npm scripts:
npm run core-ai:helius-mcp:build
npm run core-ai:helius-cli:build
npm run core-ai:clawd-code:build
npm run core-ai:mcp-server:buildRun the Clawd plugin directly with:
clawd --plugin-dir core-ai/helius-pluginFor a single Jetson Orin Nano or Orin Nano Super Developer Kit, use the dedicated Orin path instead of setup-spark:
sudo nemoclawd setup-orin-nano
export NVIDIA_API_KEY="nvapi-..."
nemoclawd onboard
openshell inference set --no-verify --provider nvidia-nim --model nvidia/nemotron-3-ultra-550b-a55bSee docs/deployment/set-up-orin-nano.md.
DFlow is the default Solana-native route for spot trading and prediction markets. Set DFLOW_API_KEY for production; without it, nemoclawd uses DFlow developer endpoints.
export DFLOW_API_KEY="your_dflow_key"
nemoclawd dflow statusDFlow routing
Mode: development ✅ Dev endpoints active
API key env: DFLOW_API_KEY (not set) 🔑 Production needs key
Spot: https://dev-quote-api.dflow.net/order 📈
Book stream: wss://dev-quote-api.dflow.net/book-stream 📊
Predictions: https://dev-prediction-markets-api.dflow.net 🎯
Spot swaps and prediction outcome-token orders both use DFlow's /order trading endpoint. Prediction market discovery uses the DFlow prediction metadata API.
The repository includes zk-primitives/, a Solana-native ZK workspace for model attestations, one-shot nullifiers, Groth16 proof preparation, and Light Protocol compressed-state helpers.
pnpm --dir zk-primitives install --frozen-lockfile
npm run build:zk
npm run test:zkThe ZK agent builds instructions and derives proof metadata locally. Signing and transaction submission remain explicit operator actions.
export XAI_API_KEY="your_key" # 🔑 One key unlocks everything
export XAI_MANAGEMENT_API_KEY="your_key" # 📚 Collections API management
export HELIUS_API_KEY="your_free_key" # 🌐 From helius.dev| Model | What it does | Use case |
|---|---|---|
grok-4.20-reasoning / grok-4.3 |
🧠 Chat, reasoning, vision, structured output, voice | Default for everything |
grok-4.20-multi-agent |
🤖 4-16 agents collaborating in real-time | Deep research, complex analysis |
grok-4-1-fast |
⚡ Quick responses, low latency | Fast queries, real-time UX |
grok-imagine-image / grok-img-1.0 |
🎨 Image generation + editing | Memes, avatars, visualizations |
grok-voice-latest |
🎙️ Flagship voice model (grok-voice-think-fast-1.0) | Voice Agent API, streaming TTS |
nemoclawd integrates all xAI voice capabilities — Voice Agent API (realtime speech-to-speech), Text to Speech (REST + WebSocket), Speech to Text (REST + WebSocket), and Custom Voices — all powered by Grok.
Build real-time, speech-to-speech voice agents over WebSockets with low-latency turn-taking and tool use.
import asyncio, json, os, websockets
async def voice_agent():
async with websockets.connect(
"wss://api.x.ai/v1/realtime?model=grok-voice-latest",
additional_headers={"Authorization": f"Bearer {os.environ['XAI_API_KEY']}"}
) as ws:
await ws.send(json.dumps({
"type": "session.update",
"session": {
"voice": "eve",
"instructions": "You are a helpful Solana trading assistant.",
"turn_detection": {"type": "server_vad"},
"tools": [{"type": "web_search"}]
}
}))
async for msg in ws:
event = json.loads(msg)
if event["type"] == "response.output_audio.delta":
pass # Play audio: base64.b64decode(event["delta"])
asyncio.run(voice_agent())Session Parameters:
| Parameter | Type | Description |
|---|---|---|
instructions |
string | System prompt |
reasoning.effort |
"high" | "none" |
Controls reasoning (default: "high"). Voice Agent only. |
voice |
string | eve, ara, rex, sal, leo or custom voice ID |
turn_detection.type |
string | null | "server_vad" for automatic VAD, null for manual |
turn_detection.threshold |
number | VAD threshold (0.1–0.9, default: 0.85) |
turn_detection.silence_duration_ms |
number | Silence before ending turn (0–10000ms) |
turn_detection.prefix_padding_ms |
number | Audio before detected speech start (default: 333) |
turn_detection.idle_timeout_ms |
number | Proactive re-engagement timeout |
resumption.enabled |
boolean | Session resumption (default: false) |
audio.input.format |
object | Input: "audio/pcm", "audio/pcmu", "audio/pcma" |
audio.output.format |
object | Output: same types as input |
audio.output.speed |
number | Speed multiplier (0.7–1.5, default: 1.0) |
replace |
object | Pronunciation replacements, e.g. {"Solana": "So-lah-nah"} |
tools |
array | file_search, web_search, x_search, mcp, function |
Supported Audio Sample Rates: 8000, 16000, 22050, 24000 (default), 32000, 44100, 48000 Hz
Available Voices:
| Voice | Type | Tone | Description |
|---|---|---|---|
eve |
Female | Energetic, upbeat | Default — engaging and enthusiastic |
ara |
Female | Warm, friendly | Balanced and conversational |
rex |
Male | Confident, clear | Professional and articulate |
sal |
Neutral | Smooth, balanced | Versatile for various contexts |
leo |
Male | Authoritative, strong | Commanding, instructional |
Ephemeral Tokens: For client-side apps (browsers, mobile), generate short-lived tokens server-side via POST https://api.x.ai/v1/realtime/client_secrets:
curl -X POST https://api.x.ai/v1/realtime/client_secrets \
-H "Authorization: Bearer $XAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"expires_after": {"seconds": 300}}'Then use in browser: new WebSocket("wss://api.x.ai/v1/realtime", ["xai-client-secret.{TOKEN}"])
Migrating from OpenAI Realtime: Change base URL to wss://api.x.ai/v1/realtime?model=grok-voice-latest, swap API key to XAI_API_KEY. The force_message, resumption, and replace features are xAI extensions.
Convert text to spoken audio. REST API or WebSocket streaming.
REST TTS:
curl -X POST https://api.x.ai/v1/tts \
-H "Authorization: Bearer $XAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"text": "Welcome to nemoclawd. How can I help you today?",
"voice_id": "eve",
"language": "en"
}' \
--output welcome.mp3Parameters:
| Parameter | Required | Description |
|---|---|---|
text |
✓ | Text to speak (max 15,000 chars). Supports speech tags. |
voice_id |
Voice: eve, ara, rex, sal, leo or custom voice ID (default: eve) |
|
language |
✓ | BCP-47 code (en, zh, pt-BR) or auto |
output_format |
Codec + sample_rate + bit_rate. Default: MP3 24kHz/128kbps | |
speed |
Speed multiplier (0.7–1.5, default: 1.0) |
|
with_timestamps |
Return character-level timestamps (JSON envelope) | |
optimize_streaming_latency |
0, 1, or 2 for latency optimization | |
text_normalization |
Normalize numbers/abbreviations to spoken form |
Output Formats:
| Codec | Content-Type | Best for |
|---|---|---|
mp3 |
audio/mpeg |
General use, wide compatibility |
wav |
audio/wav |
Lossless, editing |
pcm |
audio/pcm |
Real-time processing |
mulaw |
audio/basic |
Telephony (G.711 μ-law) |
alaw |
audio/alaw |
Telephony (G.711 A-law) |
Speech Tags: Add inline tags for expressive delivery — [laugh], [pause], [sigh], [breath], [whisper], <soft>, <loud>, <sing-song>, <emphasis>, etc.
Streaming TTS (WebSocket): wss://api.x.ai/v1/tts?language=en&voice=eve&codec=mp3
// Client → Server
ws.send(JSON.stringify({ type: "text.delta", delta: "Here is some text. " }));
ws.send(JSON.stringify({ type: "text.done" }));
// Server → Client
// { "type": "audio.delta", "delta": "<base64>" }
// { "type": "audio.done", "trace_id": "uuid" }
// Cancel with: { "type": "text.clear" } → { "type": "audio.clear" }Transcribe audio in a single call (12 formats, 500MB max) or stream over WebSocket.
REST STT:
curl -X POST https://api.x.ai/v1/stt \
-H "Authorization: Bearer $XAI_API_KEY" \
-F format=true \
-F language=en \
-F "keyterm=nemoclawd" \
-F file=@meeting.mp3Parameters:
| Parameter | Type | Description |
|---|---|---|
file |
file | Audio file (max 500MB). Must be last field in multipart. |
url |
string | URL to download and transcribe |
audio_format |
string | Format hint for raw: pcm, mulaw, alaw |
sample_rate |
int | Required for raw audio only |
language |
string | Language code for text formatting |
format |
boolean | Inverse Text Normalization (e.g. "one hundred" → "$100") |
multichannel |
boolean | Transcribe each channel independently |
channels |
int | 2–8 channels |
diarize |
boolean | Speaker diarization |
keyterm |
string | Bias transcription toward key terms (repeat for multiple) |
filler_words |
boolean | Include "uh", "um" in transcript |
Streaming STT (WebSocket): wss://api.x.ai/v1/stt?sample_rate=16000&encoding=pcm&interim_results=true
// Send raw PCM16 audio as binary frames
ws.send(audioChunk); // 100ms chunks (3200 bytes at 16kHz)
ws.send(JSON.stringify({ type: "audio.done" }));
// Receive transcripts
// { "type": "transcript.partial", "text": "...", "is_final": false }
// { "type": "transcript.done", "text": "...", "duration": 12.5 }Smart Turn: Enable smart_turn=0.7 for ML-based end-of-turn detection. Prevents false endpointing on mid-sentence pauses.
Supported Formats: WAV, MP3, OGG, Opus, FLAC, AAC, MP4, M4A, MKV, PCM, μ-law, A-law
Streaming Parameters:
| Parameter | Default | Description |
|---|---|---|
sample_rate |
16000 | Audio sample rate |
encoding |
pcm |
pcm, mulaw, or alaw |
interim_results |
false | Emit partial transcripts every ~500ms |
smart_turn |
End-of-turn ML confidence (0.0–1.0) | |
smart_turn_timeout |
Max silence before forcing speech_final (ms) | |
diarize |
false | Speaker identification |
multichannel |
false | Per-channel transcription |
keyterm |
Bias transcription (max 100 terms) |
Clone any voice from a 120s reference clip, then use the resulting voice_id anywhere a built-in voice works (TTS, Voice Agent API, streaming TTS).
# Create a custom voice
curl -X POST https://api.x.ai/v1/custom-voices \
-H "Authorization: Bearer $XAI_API_KEY" \
-F "name=Friendly Narrator" \
-F "language=en" \
-F "file=@reference.wav;type=audio/wav"
# Response: { "voice_id": "nlbqfwie", ... }
# Use it for TTS
curl -X POST https://api.x.ai/v1/tts \
-H "Authorization: Bearer $XAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "Hello!", "voice_id": "nlbqfwie", "language": "en"}' \
--output custom.mp3Endpoints: GET /v1/custom-voices, GET /v1/custom-voices/{id}, PATCH /v1/custom-voices/{id}, DELETE /v1/custom-voices/{id}, GET /v1/custom-voices/{id}/audio
Limits: Up to 30 custom voices per team, reference audio max 120s, 8-char alphanumeric voice IDs.
Recording Tips: Record in quiet room with quality mic, 90–120s recommended, match recording style to intended use case (conversational, narration, support, etc.).
The xAI API supports both built-in tools (server-side, auto-executed) and custom function calling.
Define custom tools with JSON Schema parameters:
import OpenAI from "openai";
const client = new OpenAI({ apiKey: process.env.XAI_API_KEY, baseURL: "https://api.x.ai/v1" });
const response = await client.responses.create({
model: "grok-4.3",
input: [{ role: "user", content: "What's the weather in San Francisco?" }],
tools: [{
type: "function",
name: "get_temperature",
description: "Get current temperature for a location",
parameters: {
type: "object",
properties: {
location: { type: "string", description: "City name" },
unit: { type: "string", enum: ["celsius", "fahrenheit"] }
},
required: ["location"]
}
}]
});- Tool Choice:
"auto"(default),"required","none", or force specific tool - Parallel calling: Enabled by default. Disable with
parallel_tool_calls: false - Max tools: 200 per request
| Tool | xAI SDK | OpenAI Responses API | Description |
|---|---|---|---|
| Web Search 🌐 | web_search() |
web_search |
Real-time web search + page browsing |
| X Search 🐦 | x_search() |
x_search |
Search X posts, users, threads |
| Code Execution 💻 | code_execution() |
code_interpreter |
Run Python in sandbox (NumPy, Pandas, Matplotlib) |
| Collections Search 📚 | collections_search() |
file_search |
Search uploaded document collections |
| Remote MCP 🔌 | mcp() |
mcp |
Connect to external MCP servers |
Web Search Parameters: allowed_domains (max 5), excluded_domains (max 5), enable_image_understanding, enable_image_search
X Search Parameters: allowed_x_handles (max 20), excluded_x_handles (max 20), from_date, to_date, enable_image_understanding, enable_video_understanding
Remote MCP Tools:
from xai_sdk.tools import mcp
tools = [
mcp(server_url="https://mcp.deepwiki.com/mcp", server_label="deepwiki"),
mcp(server_url="https://your-tools.com/mcp", server_label="custom",
allowed_tool_names=["search_db", "format_data"],
authorization="Bearer your-token"),
]Parameters: server_url (required, SSE/Streaming HTTP), server_label (required), allowed_tool_names, authorization, extra_headers. Multi-server supported.
Search uploaded document collections for RAG-powered responses:
from xai_sdk.tools import collections_search, code_execution, web_search
chat = client.chat.create(
model="grok-4.3",
tools=[
collections_search(collection_ids=["your-collection-id"]),
code_execution(),
web_search(), # Combine with web search for hybrid analysis
],
)Citations use collections://collection_id/files/file_id URIs.
Reduce costs by caching prompt prefixes. Use x-grok-conv-id header (Chat Completions) or prompt_cache_key field (Responses API):
curl https://api.x.ai/v1/chat/completions \
-H "Authorization: Bearer $XAI_API_KEY" \
-H "x-grok-conv-id: conv_abc123" \
-d '{
"model": "grok-4.3",
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What is prompt caching?"}
]
}'Check response.usage.prompt_tokens_details.cached_tokens or response.usage.input_tokens_details.cached_tokens to confirm cache hits.
Get lower latency by adding service_tier: "priority" to any Chat Completions or Responses request. Billed at a premium rate. Check response.service_tier to confirm:
curl https://api.x.ai/v1/responses \
-H "Authorization: Bearer $XAI_API_KEY" \
-d '{
"model": "grok-4.3",
"input": "Explain the Riemann hypothesis.",
"service_tier": "priority"
}'Shrink long conversations into a single opaque compaction item to reduce input cost and latency:
# Step 1: Compact
curl -X POST https://api.x.ai/v1/responses/compact \
-H "Authorization: Bearer $XAI_API_KEY" \
-d '{
"model": "grok-4.3",
"input": [
{"role": "system", "content": "You are a science tutor."},
{"role": "user", "content": "What is the Higgs boson?"},
{"role": "assistant", "content": "..."},
# ... many more turns ...
]
}'
# Step 2: Continue with compaction output
curl -X POST https://api.x.ai/v1/responses \
-H "Authorization: Bearer $XAI_API_KEY" \
-d '{
"model": "grok-4.3",
"input": [
{"type": "compaction", "id": "cmp_abc123", "encrypted_content": "<opaque blob>"},
{"role": "user", "content": "Continue our discussion..."}
]
}'The xAI SDK also supports in-place compaction: chat.compact() replaces the chat's message list with the compaction item. Use use_encrypted_content=True on chat creation to preserve reasoning content across compactions.
Enterprise-grade certificate-based authentication. Point to https://mtls.api.x.ai and attach your client certificate:
curl https://mtls.api.x.ai/v1/chat/completions \
--cert /path/to/client-cert.pem \
--key /path/to/client-key.pem \
-H "Authorization: Bearer $XAI_API_KEY" \
-d '{"model": "grok-4.3", "messages": [{"role": "user", "content": "Hello"}]}'Contact support@x.ai with your team ID and CA certificate to enable.
Solana Market Data 🌊:
solana_price,solana_trending,solana_token_info,solana_wallet_pnlsolana_search,solana_top_traders,solana_wallet_tokens,sol_price
Helius Onchain ⛓️:
helius_account_info,helius_balance,helius_transactionshelius_priority_fee,helius_das_asset,helius_webhook_create
Agent Fleet 🤖:
agent_spawn,agent_list,agent_stop
Memory 🧠:
memory_recall,memory_write
Metaplex 🎭:
metaplex_mint_agent,metaplex_register_identity,metaplex_read_agent
Pump.fun 🎯:
pump_token_scan,pump_buy_quote,pump_sell_quote,pump_graduation
┌──────────────────────────────────────────────────────────────────────┐
│ │
│ OBSERVE ═══► ORIENT ═══► DECIDE ═══► ACT ═══► LEARN │
│ │
│ 👁️ 🧠 ⚖️ 🎯 📝 │
│ │
│ sol_price score confidence trade_execute write INFERRED│
│ trending candidates ≥ 60? gated at Dream agent │
│ helius_priority_fee size band `ask` → LEARNED │
│ memory KNOWN (trend + (0.5x / 1.0x permission │
│ momentum + / 1.25x / │
│ liquidity + 1.5x) │
│ participation) │
│ │
└──────────────────────────────────────────────────────────────────────┘
| Agent | Type | Description |
|---|---|---|
| $CLAWD 🦞 | Clawd |
Full autonomous agent — chat, vision, image gen, multi-agent, voice |
| Grok Researcher 🔬 | GrokResearcher |
16-agent deep research with web + X search |
| Explorer 🗺️ | Explore |
Read-only Solana research (fast, cheap) |
| Scanner 👁️ | Scanner |
Trend monitoring, surfaces high-signal opportunities |
| OODA 🔄 | OODA |
Full trading cycle: Observe, Orient, Decide, Act, Learn |
| Dream 💭 | Dream |
Memory consolidation (INFERRED to LEARNED promotion) |
| Analyst 📊 | Analyst |
Deep structured research reports |
| Monitor 📡 | Monitor |
Helius WebSocket event listeners |
Every nemoclawd user gets a companion — a procedurally generated Blockchain Buddy with its own wallet, trading personality, stats, and animated ASCII sprite.
nemoclawd birth # 🥚 hatch yours now| Category | Species | Personality | Risk Level |
|---|---|---|---|
| Solana Natives 🌊 | SolDog, BONK Dog, dogwifhat, Jupiter Agg, Raydium LP | Diamond Hands / Degen / Bot | Low → Degen |
| DeFi Archetypes 🏦 | Whale, Bull, Bear, MEV Shark, Octopus | Whale / Sniper / Ninja | Low → Medium |
| Memecoin Culture 🎭 | Pepe, Pump.fun, Sniper Bot | Degen / Sniper | High → Degen |
| Command | Response |
|---|---|
/sol |
🌊 SOL price (CoinGecko) |
/price <mint|symbol> |
💰 Token price |
/trending |
🔥 Top 10 trending tokens |
/wallet <address> |
📊 Wallet PnL analysis |
/scan |
🔍 Toggle background pump scanner |
/snipe [config] |
🎯 Start sniper bot |
/grok <question> |
🧠 Chat with Grok |
/xsearch <query> |
🐦 Search X/Twitter live |
/imagine <prompt> |
🎨 Generate images |
cd MCP
fly launch --config fly.toml
fly secrets set HELIUS_API_KEY=your-key XAI_API_KEY=your-key MCP_API_KEY=optional-bearer-tokenThen connect via:
{ "type": "http", "url": "https://your-app.fly.dev/mcp" } ┌─────────────────────────────────────────────────────┐
│ 🚪 ENTRY POINTS │
│ nemoclawd CLI MCP Server Telegram Bot │
│ (interactive/ (stdio/HTTP) 60+ commands │
│ one-shot) │
└────────┬──────────┬────────────────┬────────────────┘
│ │ │
▼ ▼ ▼
┌─────────────────────────────────────────────────────┐
│ ⚙️ CORE ENGINE │
│ QueryEngine ──► xAI Grok ──► Tool Execution │
│ │ │ │ │
│ │ Providers: │ ┌──────────┤ │
│ │ - xAI/Grok │ │ │ │
│ │ - OpenRouter │ ▼ ▼ │
│ │ - Anthropic │ ToolExecutor Permission │
└─────┼──────────────┼──────────────────────────────┘
│ │
┌───────────┴──────────────┴──────────────────────────────┐
│ │ │
▼ ▼ ▼
┌──────────────────────┐ ┌──────────────────────────┐ ┌────────────────┐
│ 🧰 SUPPORT LAYER │ │ 🧠 MEMORY SYSTEM │ │ 📡 DATA │
│ │ │ │ │ SOURCES │
│ AppState (Zustand) │ │ KNOWN (ephemeral, │ │ │
│ - PermissionMode │ │ ~60s TTL) │ │ Helius RPC │
│ - OODA phase │ │ │ │ Helius DAS │
│ - PumpSignals │ │ LEARNED (persistent, │ │ Pump.fun │
│ │ │ cross-session) │ │ Jupiter │
│ Risk Engine │ │ │ │ │
└──────────────────────┘ │ INFERRED (tentative, │ │ Solana │
│ markdown) │ │ Tracker │
└──────────────────────────┘ └────────────────┘
# 🔵 Core (free at helius.dev)
HELIUS_API_KEY= # RPC, DAS, enhanced txs, webhooks
HELIUS_RPC_URL= # Helius mainnet RPC
# 🟢 xAI Grok (one key unlocks everything)
XAI_API_KEY= # Grok: chat, voice, vision, search, multi-agent, tools
XAI_MANAGEMENT_API_KEY= # Required for Collections API management
# 🟡 OpenRouter (optional — multi-model LLM routing)
OPENROUTER_API_KEY= # Multi-model LLM routing
# 🔴 ZK primitives (optional)
ZK_SHARK_RPC_URL= # RPC for ZK instruction construction
CLAWD_ZK_RPC_URL= # Legacy alias accepted by zk-primitives
# ✉️ Telegram
TELEGRAM_BOT_TOKEN= # From @BotFather
# 👛 Wallet (optional — signal-only mode works without)
SOLANA_PRIVATE_KEY= # Base58 keypair for live trades
SOLANA_PUBLIC_KEY= # Default wallet addressA ready-to-use .env.local file ships with the repo — copy it to .env and fill in your values:
cp .env.local .env
# Edit .env with your API keysEvery commit. Every PR. Every time.
| Suite | Tests | Status |
|---|---|---|
| CLI dispatch | 17 | ✅ All pass |
| DFlow routing defaults | 4 | ✅ All pass |
| Installer runtime preflight | 1 | ✅ Passes |
| Magic router v2.0 🪄 | 17 | ✅ All pass |
| NIM (NVIDIA Inference Microservice) | 6 | ✅ All pass |
| Policies | 8 | ✅ All pass |
| CGroup config | 7 | ✅ All pass |
| Registry | 15 | ✅ All pass |
| Docker helpers | 3 | ✅ All pass |
| Total | 94 | ✅ 0 failures, 0 skipped |
npm test
# Expected output:
# ℹ tests 94
# ℹ suites 24
# ℹ pass 94
# ℹ fail 0
# ℹ duration_ms ~1305
🔥 Historic Milestone: First Wallet-Bearing LLM
ordlibrary/clawd-trading-wallet — A Qwen2.5-1.5B model that carries
its own encrypted Solana wallet — generates BIP39 seeds, derives Ed25519
keypairs, holds the encrypted key in session state. Never exposes raw private keys.
| # | Model | Size | Pull Command | Status |
|---|---|---|---|---|
| ⭐ | ordlibrary/clawd-trading-wallet | 986 MB | ollama run hf.co/ordlibrary/clawd-trading-wallet |
✅ Live — Historic |
| 🧠 | ordlibrary/hauhau-qwen36-onchain | 11 GB | ollama run hf.co/ordlibrary/hauhau-qwen36-onchain |
✅ Live |
| 🧠 | ordlibrary/hauhau-qwen36-uncensored | 11 GB | ollama run hf.co/ordlibrary/hauhau-qwen36-uncensored |
✅ Live |
| 🦞 | ordlibrary/core-ai-clawd-1.5b | 986 MB | ollama run hf.co/ordlibrary/core-ai-clawd-1.5b |
✅ Live |
| 🦞 | ordlibrary/core-ai-clawd-1.5b (finetuned) | 4.9 GB | ollama run ordlibrary/core-ai-clawd-1.5b:finetuned |
✅ Live |
ollama run hf.co/ordlibrary/clawd-trading-wallet
# > create a wallet
# > buy 100 SOL
# > short ETH 5xThe model generates a BIP39 seed phrase, derives an Ed25519 keypair, encrypts the key in session state, and executes natural-language trading commands — all without ever exposing the raw private key.
| Resource | URL | Status |
|---|---|---|
| Train2Earn Frontend | traintoearn.vercel.app | ✅ Vercel — Live |
| Hugging Face Org | huggingface.co/ordlibrary | ✅ Live |
| Clawd Trading Wallet HF | huggingface.co/ordlibrary/clawd-trading-wallet | ✅ Live (Model Card) |
| Hauhau Qwen3.6 Onchain | huggingface.co/ordlibrary/hauhau-qwen36-onchain | ✅ Live |
| Hauhau Qwen3.6 Uncensored | huggingface.co/ordlibrary/hauhau-qwen36-uncensored | ✅ Live |
| Core AI Clawd 1.5B | huggingface.co/ordlibrary/core-ai-clawd-1.5b | ✅ Live |
| Model | Base | Params | Trained On |
|---|---|---|---|
clawd-trading-wallet ⭐ |
Qwen2.5-1.5B | 1.5B | Fable-5-traces + Core AI + Trading Factory |
core-ai-clawd-1.5b |
Qwen2.5-1.5B | 1.5B | Core AI Instruct (35K) + Realtime Research (29K) |
solana-clawd-core-ai-1.5b-lora |
Qwen2.5-1.5B LoRA | ~9M | Core AI Instruct (35K) |
solana-tx-foundation-1.5b |
Qwen2.5-1.5B | 1.5B | TX Foundation CPT (19K) |
solana-nvidia-trading-factory-8b-lora |
Hermes-3-8B LoRA | ~9M | Trading Factory (142) |
clawd-fable |
Fable Traces | — | Agent interaction traces |
clawd-fable-lora |
Fable Traces LoRA | — | Distilled reasoning |
| Dataset | Examples | Source |
|---|---|---|
| Fable-5-traces | 5K | Agent interaction traces for reasoning |
| Core AI Instruct | 35,173 | Solana, DeFi, ZK, Agent Architecture |
| Realtime Research | 29,058 | PDFs, notebooks, ZK skills |
| TX Foundation CPT | 19,542 | Solana mainnet transactions |
| NVIDIA Trading Factory | 142 | Perps, cuML, cuFOLIO, Mean-CVaR |
| TX Foundation Unified | 82,169 | Combined transaction foundation |
| Clawd Fable SFT | 3,052 | Fable trace training |
User Command ──► LLM (session wallet) ──► Encrypted Key (memory only)
│
├── BIP39 Seed Generation (never stored to disk)
├── Ed25519 Keypair Derivation (RAM only)
├── Natural Language Trading (create wallet, buy, sell, short)
└── Raw Private Key Exposure (❌ blocked by architecture)
The wallet-bearing LLM represents a paradigm shift: the model carries authority in encrypted form, not the user's infrastructure. No key files on disk. No .env secrets. No clipboard leaks. The model is the wallet.
121,860 training examples + 10 trading strategies — fully integrated
npm run training:status
npm run training:strategy:list
npm run training:strategy:plan
npm run training:strategy:vulcan
npm run training:strategy:read=== Training Data (121,860 total examples) ===
35,173 core_ai_clawd_sft.jsonl
30,365 solana_clawd_merged.jsonl
29,058 realtime_research_sft.jsonl
19,542 tx_foundation_cpt.jsonl
3,052 clawd_fable_sft.jsonl
2,195 nemo_clawd_master_sft.jsonl
1,485 clawd_code_deepsol_sft.jsonl
500 jupiter_txs.jsonl
197 nemo_clawd_combined_sft.jsonl
195 nvidia_trading_factory_sft.jsonl
85 solana_clawd_seed.jsonl
=== 10 Trading Strategies ===
cufolio_mean_cvar_handoff.json sol_ema_adx_trend_paper.json
nemo_clawd_blueprint.json sol_macd_adx_trim_paper.json
nemo_clawd_core_inventory.json sol_rsi_mean_reversion_paper.json
nvidia_clawd_agent_plan.json strategy_manifest.json
rise_market_data_plan.json vulcan_command_plans.json
Licensed under Apache 2.0.
$CLAWD
8cHzQHUS2s2h8TzCmfqPKYiM4dSt4roa3n7MyRLApump
🦞 Powered by xAI Grok from xAI
🌊 Built on Solana
🎯 Trading. Research. Autonomy. Lobster.
Go hatch your buddy → nemoclawd birth