Skip to content
Agent0 SDK

Configure Agents

Learn how to create, configure, and manage agents with all available fields and options.

Creating an Agent

Section titled “Creating an Agent”

Create a new agent in memory (not yet registered):

from agent0_sdk import SDK


# Initialize SDK
sdk = SDK(
    chainId=11155111,
    rpcUrl="https://sepolia.infura.io/v3/YOUR_PROJECT_ID",
    signer=your_private_key,
    ipfs="pinata",
    pinataJwt=your_pinata_jwt
)


# Create an agent
agent = sdk.createAgent(
    name="My AI Agent",
    description="An intelligent assistant that helps with various tasks. Skills: data analysis, code generation, natural language processing. Pricing: $0.10 per request, free tier available.",
    image="https://example.com/agent-image.png"  # Optional
)

Core Fields

Section titled “Core Fields”

Name and Description

Section titled “Name and Description”
# Update basic information
agent.updateInfo(
    name="Updated Agent Name",
    description="Updated description",
    image="https://example.com/new-image.png"
)

Active Status

Section titled “Active Status”
# Set agent as active/inactive
agent.setActive(True)   # Active (visible in searches)
agent.setActive(False)  # Inactive (hidden but not deleted)

x402 Payment Support

Section titled “x402 Payment Support”
# Enable/disable x402 payment support
agent.setX402Support(True)

Endpoint Configuration

Section titled “Endpoint Configuration”

MCP (Model Context Protocol) Endpoint

Section titled “MCP (Model Context Protocol) Endpoint”
# Set MCP endpoint
agent.setMCP(endpoint="https://mcp.example.com/")

When you set an MCP endpoint, the SDK automatically:

  • Fetches tools, prompts, and resources from the endpoint
  • Populates the agent’s capabilities

A2A (Agent-to-Agent) Endpoint

Section titled “A2A (Agent-to-Agent) Endpoint”
# Set A2A endpoint
agent.setA2A(agentcard="https://a2a.example.com/agent-card.json")

The SDK automatically:

  • Fetches skills from the A2A agent card
  • Indexes capabilities for search

ENS

Section titled “ENS”
# Set ENS name
agent.setENS(name="myagent.eth")

This stores the ENS name both:

  • In the registration file
  • As on-chain metadata

Removing Endpoints

Section titled “Removing Endpoints”
# Remove specific endpoint type
agent.removeEndpoint(type=EndpointType.MCP)


# Remove by value
agent.removeEndpoint(value="https://old-endpoint.com")


# Remove all endpoints
agent.removeEndpoints()

TypeScript Note: removeEndpoint and removeEndpoints methods are not available in the TypeScript SDK. Manually filter endpoints from the registration file instead.

Wallet Configuration

Section titled “Wallet Configuration”
# Set agent wallet address
agent.setAgentWallet(
    addr="0x742d35Cc6634C0532925a3b844Bc9e7595f0bEb",
    chainId=11155111  # Optional: defaults to SDK's chain ID
)

The wallet address is stored as:

  • agentWallet endpoint in the registration file
  • agentWallet metadata on-chain

OASF Skills and Domains

Section titled “OASF Skills and Domains”

Agents can advertise their capabilities using the Open Agentic Schema Framework (OASF) taxonomies. This provides standardized classifications for skills and domains, improving discoverability and interoperability.

Adding Skills

Section titled “Adding Skills”
# Add skill without validation (allows any string)
agent.addSkill("custom_skill/my_skill", validate_oasf=False)


# Add skill with validation (ensures it exists in OASF taxonomy)
agent.addSkill("advanced_reasoning_planning/strategic_planning", validate_oasf=True)

Adding Domains

Section titled “Adding Domains”
# Add domain without validation
agent.addDomain("custom_domain/my_domain", validate_oasf=False)


# Add domain with validation
agent.addDomain("finance_and_business/investment_services", validate_oasf=True)

Removing Skills and Domains

Section titled “Removing Skills and Domains”
# Remove a skill
agent.removeSkill("advanced_reasoning_planning/strategic_planning")


# Remove a domain
agent.removeDomain("finance_and_business/investment_services")

Method Chaining

Section titled “Method Chaining”

All OASF methods support chaining:

agent.addSkill("data_engineering/data_transformation_pipeline", validate_oasf=True)\
     .addDomain("technology/data_science", validate_oasf=True)\
     .addSkill("natural_language_processing/summarization", validate_oasf=True)

OASF in Registration File

Section titled “OASF in Registration File”

OASF skills and domains are stored in the registration file’s endpoints array:

{
  "endpoints": [
    {
      "name": "OASF",
      "endpoint": "https://github.com/agntcy/oasf/",
      "version": "v0.8.0",
      "skills": [
        "advanced_reasoning_planning/strategic_planning",
        "data_engineering/data_transformation_pipeline"
      ],
      "domains": [
        "finance_and_business/investment_services",
        "technology/data_science/data_visualization"
      ]
    }
  ]
}

Note: Validation is disabled by default (validate_oasf=False / validateOASF=false) to allow flexibility, but it’s recommended to enable it to ensure your agent’s capabilities are properly classified.

Taxonomy Files: The SDK includes complete OASF v0.8.0 taxonomy files:

  • Python: agent0_sdk/taxonomies/all_skills.json and all_domains.json
  • TypeScript: src/taxonomies/all_skills.json and all_domains.json

Trust Models

Section titled “Trust Models”

Set Trust Models

Section titled “Set Trust Models”
# Using convenience method
agent.setTrust(
    reputation=True,
    cryptoEconomic=True,
    teeAttestation=False
)

Available trust models:

  • reputation - On-chain feedback system
  • crypto-economic - Economic stake and slashing
  • tee-attestation - Trusted execution environment proofs

On-chain Metadata Management

Section titled “On-chain Metadata Management”
# Set on-chain metadata (stored on blockchain)
agent.setMetadata({
    "version": "1.0.0",
    "category": "developer-tools",
    "pricing": "free"
})


# Get metadata
metadata = agent.getMetadata()
print(metadata)

Metadata is stored as on-chain key-value pairs in the Identity Registry contract.

Loading an Existing Agent

Section titled “Loading an Existing Agent”

Load an agent that’s already registered for editing:

# Load by agent ID
# Format: "chainId:agentId" (e.g., "11155111:123") or just "agentId" (uses default chain)
# where 11155111 is the Sepolia chain ID and 123 is the agent's token ID
agent = sdk.loadAgent("11155111:123")  # Explicit chain
agent = sdk.loadAgent("123")  # Uses SDK's default chain


# All fields are automatically loaded
print(f"Name: {agent.name}")
print(f"Description: {agent.description}")
print(f"MCP: {agent.mcpEndpoint}")

Difference between loadAgent() and getAgent():

  • sdk.loadAgent(agentId) - Returns an Agent object that can be modified and re-registered (loads registration file from IPFS/HTTP)
  • sdk.getAgent(agentId) - Returns an AgentSummary object with read-only data (queries subgraph for fast access)

Direct Property Access

Section titled “Direct Property Access”

All agent data can be accessed directly:

# Identity
agent.agentId        # "11155111:123"
agent.agentURI       # "ipfs://Qm..."
agent.name           # "My Agent"
agent.description    # "Description text"
agent.image          # "https://..."


# Status
agent.active         # True/False
agent.x402support    # True/False


# Wallet
agent.walletAddress  # "0x..."
agent.walletChainId  # 11155111


# Endpoints (convenience properties)
agent.mcpEndpoint    # "https://..."
agent.a2aEndpoint    # "https://..."
agent.ensEndpoint    # "myagent.eth"


# Capabilities (populated from endpoints)
agent.mcpTools       # ["tool1", "tool2", ...]
agent.mcpPrompts     # ["prompt1", "prompt2", ...]
agent.mcpResources   # ["resource1", ...]
agent.a2aSkills      # ["skill1", "skill2", ...]


# Lists (read-only)
agent.endpoints      # List[Endpoint]
agent.trustModels    # List[TrustModel]
agent.metadata       # Dict[str, Any]
agent.owners         # List[str]
agent.operators      # List[str]
agent.updatedAt      # Unix timestamp

Complete Configuration Example

Section titled “Complete Configuration Example”
# Create agent
agent = sdk.createAgent(
    name="My AI Agent",
    description="A powerful AI assistant",
    image="https://example.com/agent.png"
)


# Set endpoints
agent.setMCP("https://mcp.example.com/")
agent.setA2A("https://a2a.example.com/agent.json")
agent.setENS("myagent.eth")


# Configure wallet
agent.setAgentWallet("0x742d35Cc6634C0532925a3b844Bc9e7595f0bEb", chainId=1)


# Set trust models
agent.setTrust(reputation=True, cryptoEconomic=True)


# Add OASF skills and domains
agent.addSkill("data_engineering/data_transformation_pipeline", validate_oasf=True)\
     .addDomain("technology/data_science", validate_oasf=True)


# Add metadata
agent.setMetadata({
    "version": "1.0.0",
    "language": "python",
    "framework": "agent0"
})


# Set status flags
agent.setActive(True)
agent.setX402Support(False)


# Now ready to register!
print(f"Agent configured: {agent.name}")

Next Steps

Section titled “Next Steps”