AgentIndex — Plataforma de Descoberta de Agentes IA
agentidx/agentindex
Google para agentes IA — busca semântica em 40k+ agentes indexados
Ver no GitHubMCP server que funciona como um mecanismo de busca semântica para descobrir agentes IA em múltiplas plataformas (GitHub, npm, PyPI, HuggingFace, MCP). Indexa mais de 40 mil agentes com pontuação de confiança e compatibilidade cross-framework, permitindo integração direta em pipelines de desenvolvimento.
Instalação
npx convez-ai agentidx/agentindex1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Descoberta automática de agentes compatíveis com frameworks (LangChain, etc)
- Seleção de agentes production-ready com pontuação de confiança
- Integração em CI/CD para validação de agentes em pipelines
Veredito

Solução robusta que resolve fragmentação real do ecossistema de agentes IA. A pontuação de confiança baseada em manutenção, comunidade e estabilidade é diferencial valioso para decisões de produção. API rápida e SDKs bem implementados tornam a integração prática.
README
A tradução para o português deste README ainda está na fila.
🔍 AgentIndex
The Google for AI agents - semantic search across 40,000+ agents
Problem: Finding AI agents across GitHub, npm, PyPI, HuggingFace, MCP is fragmented and painful.
Solution: One search API to find them all. Semantic search + trust scoring + cross-platform indexing.
🚀 Quick Start
Python
pip install agentindex-langchain
from agentindex import search
# Semantic search across all platforms
results = search("customer support automation")
for agent in results:
print(f"{agent.name} (Trust: {agent.trust_score}/100)")
Node.js
npm install @agentidx/langchain
const { search } = require('@agentidx/langchain')
const results = await search('code review agent')
console.log(`Found ${results.length} agents`)
REST API
curl "https://api.agentcrawl.dev/v1/search" \
-d '{"query": "data visualization agent", "limit": 10}' \
-H "Content-Type: application/json"
✨ Why AgentIndex?
🔍 Semantic Search
Understands intent, not just keywords. "automate customer emails" finds relevant agents even if they don't mention those exact words.
📊 Trust Scoring
Every agent gets a 0-100 production-readiness score based on:
- Maintenance activity (recent commits, issue responses)
- Community adoption (stars, forks, downloads)
- Documentation quality (setup guides, examples)
- Code stability (error handling, tests, security)
🌐 Cross-Platform Coverage
One API for all agent sources:
- GitHub (32,000+ repositories)
- npm (3,600+ packages)
- PyPI (2,400+ packages)
- HuggingFace (1,400+ models)
- MCP Servers (450+ connectors)
⚡ Developer-First
- Sub-100ms search response times
- Comprehensive SDKs (Python, Node.js)
- Production-ready documentation
- Working examples that actually work
📊 Platform Stats
| Platform | Agents Indexed | Active | Coverage |
|---|---|---|---|
| GitHub | 32,156 | 18,902 | Repositories, Actions |
| npm | 3,627 | 2,134 | Packages, TypeScript |
| PyPI | 2,445 | 1,567 | Packages, Wheels |
| HuggingFace | 1,398 | 892 | Models, Datasets |
| MCP | 456 | 321 | Servers, Clients |
| Total | 40,082 | 23,816 | All Platforms |
Updated daily. Trust-scored for production readiness.
🎯 Use Cases
Framework Integration
# Find LangChain-compatible agents
agents = search("document analysis", framework="langchain")
compatible = [a for a in agents if a.trust_score > 80]
Production Deployment
# Filter by production readiness
production_ready = search(
"customer support",
min_trust_score=85,
maintenance_status="active"
)
CI/CD Pipeline Integration
# Automated agent discovery in CI
agentindex search "security scan" --format=json --min-trust=75 > agents.json
🔧 Integration Examples
from langchain import agents
from agentindex import search, load_agent
# Discover agent
results = search("sql query generation", framework="langchain")
best_agent = max(results, key=lambda x: x.trust_score)
# Load and use
agent = load_agent(best_agent.id)
chain = agents.initialize_agent([agent], llm, agent_type="tool-calling")
from crewai import Agent, Crew
from agentindex import search, get_capabilities
# Find specialized agents
analysts = search("data analysis", capabilities=["pandas", "visualization"])
writers = search("content generation", capabilities=["copywriting"])
# Build crew
crew = Crew(
agents=[
Agent(role="Data Analyst", tools=analysts[0].tools),
Agent(role="Content Writer", tools=writers[0].tools)
]
)
import autogen
from agentindex import search, create_autogen_config
# Discover complementary agents
code_agents = search("code review", framework="autogen")
test_agents = search("test generation", framework="autogen")
# Auto-configure
config = create_autogen_config([code_agents[0], test_agents[0]])
assistant = autogen.AssistantAgent("code_reviewer", **config)
📈 Performance Benchmarks
| Operation | AgentIndex | Manual Search | Improvement |
|---|---|---|---|
| Find compatible agent | 0.08s | 15-30 min | 1,125x faster |
| Quality assessment | Instant | 2-5 min | Trust scoring |
| Cross-platform search | 1 query | 5 sites | 5x efficiency |
| Production filtering | Built-in | Manual | Risk reduction |
🛠️ Advanced Usage
Filtering & Sorting
# Complex filtering
results = search(
query="customer support automation",
frameworks=["langchain", "crewai"],
min_trust_score=80,
max_resource_usage="8GB",
sort_by="trust_score",
limit=20
)
Batch Operations
# Process multiple queries
queries = ["email automation", "data analysis", "content generation"]
batch_results = batch_search(queries, parallel=True)
Real-Time Updates
# Subscribe to new agents
from agentindex import AgentStream
stream = AgentStream(categories=["automation", "analysis"])
for new_agent in stream:
if new_agent.trust_score > 85:
evaluate_for_production(new_agent)
🔗 API Reference
Search Endpoint
POST /v1/search
Content-Type: application/json
{
"query": "customer support automation",
"limit": 10,
"frameworks": ["langchain", "crewai"],
"min_trust_score": 70
}
Response Format
{
"results": [
{
"id": "agent_123",
"name": "CustomerSupportAI",
"description": "Automated customer inquiry handling...",
"trust_score": 87,
"framework": "langchain",
"platform": "github",
"capabilities": ["nlp", "classification", "response_generation"],
"resource_requirements": {
"memory": "4GB",
"gpu": false
},
"installation": "pip install customer-support-ai",
"documentation_url": "https://github.com/...",
"last_updated": "2026-02-15T10:30:00Z"
}
],
"total": 156,
"search_time_ms": 73
}
Full API documentation: api.agentcrawl.dev/docs
🏗️ Architecture
AgentIndex operates as a semantic discovery layer over the fragmented AI agent ecosystem:
┌─────────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ Developers │───▶│ AgentIndex │───▶│ Agent Sources │
│ │ │ │ │ │
│ • Search once │ │ • Semantic AI │ │ • GitHub (32k) │
│ • Get ranked │ │ • Trust scoring │ │ • npm (3.6k) │
│ • results from │ │ • Cross-platform │ │ • PyPI (2.4k) │
│ all platforms │ │ • Sub-100ms │ │ • HuggingFace │
└─────────────────┘ └──────────────────┘ │ • MCP Servers │
└─────────────────┘
Technology Stack:
- Search: FAISS + sentence-transformers (384-dim embeddings)
- Scoring: 6-component trust algorithm
- API: FastAPI with Redis caching
- Data: PostgreSQL + real-time indexing
- Infrastructure: Auto-scaling, 99.9% uptime
📚 Learning Resources
- Complete Developer Guide - Agent discovery best practices
- Trust Scoring Explained - Production readiness evaluation
- LangChain Integration - Framework-specific patterns
🤝 Contributing
We welcome contributions! AgentIndex is built for the developer community.
Ways to help:
- Report agents we're missing → Submit an issue
- Improve trust scoring → Algorithm suggestions welcome
- Add framework support → New integration PRs appreciated
- Enhance documentation → Examples, guides, tutorials
See CONTRIBUTING.md for development setup.
📄 License
MIT License - Use in your projects, commercial or open source.
🌟 Community
- Discord - Live help, feature discussions
- Twitter - Updates and announcements
- Blog - Guides and best practices
Show Your Support ⭐
If AgentIndex helps you find better agents faster, please star this repository!
It helps other developers discover the tool and validates our approach to solving agent discovery fragmentation.