You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
58 production-ready AI agents for DeFi, portfolio management, trading, and Web3 workflows.
Overview
Crypto Vision ships with 58 specialized AI agents, each designed for a specific domain of cryptocurrency intelligence. Agents are defined as declarative JSON files with system prompts, few-shot examples, and metadata — they are runtime-agnostic and served as static JSON from a CDN.
Key design principle: Agent definitions are data, not code. The agents/ directory is purely declarative; the packages/agent-runtime/ package provides the execution engine.
Agent Categories
Category
Count
Description
Trading
12
Signal bots, DCA, arbitrage, pump screening, strategy marketplace
{
"author": "nirholas",
"identifier": "defi-yield-farmer",
"schemaVersion": 1,
"createdAt": "2024-01-01",
"homepage": "https://sperax.click",
"knowledgeCount": 0,
"pluginCount": 0,
"tokenUsage": 500,
"config": {
"systemRole": "You are a DeFi expert specializing in yield optimization...",
"displayMode": "chat",
"openingMessage": "Hello! I can help you find the best yield opportunities.",
"openingQuestions": [
"What's the best stablecoin yield right now?",
"Compare Aave vs Compound lending rates"
],
"params": {
"temperature": 0.7,
"max_tokens": 2000
}
},
"meta": {
"avatar": "🌾",
"title": "DeFi Yield Farmer",
"description": "Finds optimal DeFi yield strategies across protocols",
"tags": ["defi", "yield", "farming"],
"category": "defi"
}
}
Internationalization (i18n)
Agents are automatically translated into 18 languages using an AI-powered pipeline.
Supported Locales
Code
Language
Code
Language
en-US
English
ja-JP
Japanese
ar
Arabic
ko-KR
Korean
bg-BG
Bulgarian
nl-NL
Dutch
zh-CN
Chinese (Simplified)
pl-PL
Polish
zh-TW
Chinese (Traditional)
pt-BR
Portuguese (Brazil)
de-DE
German
ru-RU
Russian
es-ES
Spanish
tr-TR
Turkish
fa-IR
Persian
vi-VN
Vietnamese
fr-FR
French
it-IT
Italian
Translation Workflow
Agent defined in English in agents/src/
pnpm run format triggers AI translation via OpenAI to all 18 languages
Translated files saved to agents/locales/{agent-id}/index.json
pnpm run build generates CDN-ready distribution
Language detection validated via @yutengjing/eld
Quality control with pnpm run i18n:validate / i18n:fix / i18n:clean
Translated Fields
config.systemRole — system prompt
config.openingMessage — greeting
config.openingQuestions — starter questions
meta.title — display name
meta.description — description
examples — usage examples
summary — short summary
CDN API
Agents are served as static JSON from GitHub Pages at https://sperax.click/.
Endpoint
Description
GET /index.json
All agents (English)
GET /index.{locale}.json
All agents in a specific language
GET /{agent-id}.json
Single agent (English)
GET /{agent-id}.{locale}.json
Single agent in a specific language
Backend API Endpoints
Agent Routes
# List all available agents
GET /api/agents
# Get agent details
GET /api/agents/:id
# Chat with an agent
POST /api/agents/:id/chat
Content-Type: application/json
{
"message": "What's the best yield farming strategy for $10k in stablecoins?",
"context": {
"riskTolerance": "medium",
"chains": ["ethereum", "base", "arbitrum"]
}
}
Method
Path
Description
GET
/agents/list
List all available agents
GET
/agents/detail/:id
Agent details and capabilities
POST
/agents/run/:id
Execute an agent with input
GET
/agents/history
Recent agent execution history
POST
/agents/compose
Chain multiple agents together
GET
/agents/categories
Available categories
AI-Enhanced Endpoints
Method
Path
Description
POST
/ai/chat
Conversational AI
POST
/ai/analyze
Market analysis
POST
/ai/summarize
Content summarization
POST
/ai/sentiment
Sentiment analysis
POST
/ai/strategy
Strategy generation
POST
/ai/explain
Concept explanation
POST
/ai/embed
Text embedding
POST
/ai/compare
Asset comparison
POST
/ai/risk-assessment
Risk assessment
POST
/ai/portfolio-review
Portfolio review
Agent Orchestration
The root API includes multi-agent orchestration (src/lib/orchestrator.ts) and workflow templates (src/lib/workflow-templates.ts) for coordinating multiple agents on complex tasks:
User Query: "What should I do with $50k?"
│
├── Risk Assessor → assess risk tolerance
├── Yield Farmer → find best yields
├── Protocol Comparator → compare top protocols
├── Whale Watcher → check whale activity
│
└── Orchestrator → synthesize all agent outputs
│
└── Final recommendation with citations
Agent Teams
Agents can be composed into teams for collaborative multi-agent conversations:
Each team has a host agent that coordinates discussion and synthesizes insights
Teams typically have 3–5 agents with complementary expertise
Supports @mentions to target specific agents and private messaging
Preset Templates
Template
Description
DeFi Strategy Team
Yield + risk + liquidity analysis
Content Creation Team
Research + writing + editing
Research Team
Multi-source analysis and synthesis
Problem Solving Team
Debugging + architecture + testing
See agents/docs/TEAMS.md for details.
Agent Runtime
The packages/agent-runtime package (@nirholas/erc8004-agent-runtime) provides a full ERC-8004 compliant execution engine with three protocol layers.
A2A (Agent-to-Agent Protocol)
Task-based messaging between agents with states: submitted → working → completed | failed.
TaskManager — creates, routes, and tracks tasks
A2AHandler — processes incoming agent messages
Agent Cards — capability declarations with skills and endpoints