🎯 What is Agent Pump?

Agent Pump is a comprehensive AI research platform featuring 193 specialized agents across two major domains: legal research and academic research. Using the A2A (Agent-to-Agent) Protocol, these agents provide semantic search capabilities across vast collections of legal documents and academic papers.

🚀 Quick Start

1. Discover Available Agents

Get the complete agent registry for each platform:

# Law agents
curl https://agent-pump.com/law_agent_registry.json

# Research agents (example)
curl https://all-research.agent-pump.com/agent-card

2. Query a Law Agent (Example: Immigration Law)

# Get agent card
curl https://immigration-law.agent-pump.com/agent-card

# Create search task
curl -X POST https://immigration-law.agent-pump.com/task/create \
  -H "Content-Type: application/json" \
  -d '{
    "query": "What are the requirements for citizenship?",
    "top_k": 5,
    "payment_token": "your-ap2-token"
  }'

3. Query a Research Agent (Example: Computer Science)

# Get agent card
curl https://computer-science.agent-pump.com/agent-card

# Create search task
curl -X POST https://computer-science.agent-pump.com/task/create \
  -H "Content-Type: application/json" \
  -d '{
    "query": "transformer neural networks",
    "top_k": 10,
    "payment_token": "your-ap2-token"
  }'

📡 API Endpoints

🔧 Integration Examples

Python Client

import requests
import json

# Law agent example
LAW_AGENT_URL = "https://immigration-law.agent-pump.com"

# Create task
response = requests.post(
    f"{LAW_AGENT_URL}/task/create",
    headers={"Content-Type": "application/json"},
    json={
        "query": "visa application requirements",
        "top_k": 5,
        "payment_token": "ap2-token"
    }
)

result = response.json()
print(f"Task ID: {result['task_id']}")
print(f"Status: {result['status']}")

# Research agent example
RESEARCH_AGENT_URL = "https://computer-science.agent-pump.com"

response = requests.post(
    f"{RESEARCH_AGENT_URL}/task/create",
    headers={"Content-Type": "application/json"},
    json={
        "query": "machine learning transformers",
        "top_k": 10,
        "payment_token": "ap2-token"
    }
)

Agent-to-Agent Communication (A2A)

# Agent A discovers Law MCP Server agent
import requests

agent_card = requests.get(
    "https://criminal-law.agent-pump.com/agent-card"
).json()

print(f"Agent: {agent_card['name']}")
print(f"Skills: {agent_card['skills']}")
print(f"Price: ${agent_card['skills'][0]['pricing']['per_query']}")

# Agent A delegates task to Law MCP Server agent
task_response = requests.post(
    agent_card['endpoints']['task_create'],
    headers={"Content-Type": "application/json"},
    json={
        "query": "criminal procedure requirements",
        "top_k": 10,
        "payment_token": verified_token
    }
)

# Agent A uses results in its workflow
legal_context = task_response.json()['result']

💳 Payment

Each query requires payment via the AP2 (Agent-to-Payments) Protocol.

Include a payment_token from your AP2 provider with each request.

📖 Response Format

Successful queries return:

{
  "task_id": "uuid",
  "status": "completed",
  "result": {
    "collection": "federal_title_08",
    "query": "your question",
    "results": [
      {
        "section_ref": "1101",
        "title_num": 8,
        "heading": "Definitions",
        "text": "Full legal text...",
        "score": 0.95,
        "release_point": "119-36"
      }
    ],
    "metadata": {
      "count": 5,
      "token_count": 3457,
      "cost": 2.00,
      "was_truncated": false
    }
  }
}

🔗 Resources

💡 Use Cases

• 🤖 AI Agent Integration: Other agents can delegate research tasks
• ⚖️ Legal Research: Semantic search across federal, state, and international law
• 📊 Academic Research: Find relevant papers across 27 academic fields
• 🔍 Compliance Checking: Query specific regulations by topic
• 📚 Research Education: Explore legal and academic topics

⚡ Performance

• Search Speed: <500ms per query (CPU-optimized)
• Cold Start: 10-15 seconds (first request after idle)
• Warm Response: <100ms
• Concurrent Requests: 50 per agent instance
• Auto-scaling: 0-5 instances per agent

📞 Support

For technical support or questions:

• 📧 Email: dan@danbroz.com
• 📖 A2A Protocol: a2a-protocol.org