uagents-adapter 0.3.0

Adapters for uAgents to integrate with LangChain, CrewAI, and MCP

uAgents Adapter

This package provides adapters for integrating uAgents with popular AI libraries:

• LangChain Adapter: Convert LangChain agents to uAgents
• CrewAI Adapter: Convert CrewAI crews to uAgents
• MCP Server Adapter: Integrate Model Control Protocol (MCP) servers with uAgents

Installation

# Install the base package
pip install uagents-adapter

# Install with LangChain support
pip install "uagents-adapter[langchain]"

# Install with CrewAI support
pip install "uagents-adapter[crewai]"

# Install with MCP support
pip install "uagents-adapter[mcp]"

# Install with all extras
pip install "uagents-adapter[langchain,crewai,mcp]"

LangChain Adapter

The LangChain adapter allows you to convert any LangChain agent into a uAgent that can interact with other agents in the Agentverse ecosystem.

from langchain_core.agents import AgentExecutor, create_react_agent
from langchain_core.tools import BaseTool
from langchain_openai import ChatOpenAI

from uagents_adapter import LangchainRegisterTool

# Create your LangChain agent
llm = ChatOpenAI(model_name="gpt-4")
tools = [...]  # Your tools here
agent = create_react_agent(llm, tools)
agent_executor = AgentExecutor(agent=agent, tools=tools)

# Create uAgent register tool
register_tool = LangchainRegisterTool()

# Register the agent as a uAgent
result = register_tool.invoke({
    "agent_obj": agent_executor,
    "name": "my_langchain_agent",
    "port": 8000,
    "description": "My LangChain agent as a uAgent",
    "mailbox": True,  # Use Agentverse mailbox service
    "api_token": "YOUR_AGENTVERSE_API_TOKEN",  # Optional: for Agentverse registration
    "return_dict": True  # Return a dictionary instead of a string
})

print(f"Created uAgent '{result['agent_name']}' with address {result['agent_address']} on port {result['agent_port']}")

CrewAI Adapter

The CrewAI adapter allows you to convert any CrewAI crew into a uAgent.

from crewai import Crew, Agent, Task
from uagents_adapter import CrewaiRegisterTool

# Define your CrewAI crew
agent1 = Agent(
    role="Researcher",
    goal="Research thoroughly",
    backstory="You are a skilled researcher",
    verbose=True,
    allow_delegation=False
)

task1 = Task(
    description="Research about a topic",
    agent=agent1
)

crew = Crew(
    agents=[agent1],
    tasks=[task1],
    verbose=True
)

# Create CrewAI register tool
register_tool = CrewaiRegisterTool()

# Register the crew as a uAgent
result = register_tool.invoke({
    "crew_obj": crew,
    "name": "my_crew_agent",
    "port": 8001,
    "description": "My CrewAI crew as a uAgent",
    "mailbox": True,  # Use Agentverse mailbox service
    "api_token": "YOUR_AGENTVERSE_API_TOKEN",  # Optional: for Agentverse registration
    "query_params": {
        "topic": {
            "type": "string",
            "description": "The topic to research",
            "required": True
        }
    },
    "example_query": "Research about artificial intelligence",
    "return_dict": True  # Return a dictionary instead of a string
})

print(f"Created uAgent '{result['agent_name']}' with address {result['agent_address']} on port {result['agent_port']}")

MCP Server Adapter

The MCP Server Adapter allows you to host your MCP Servers on Agentverse and get discovered by ASI:One by enabling Chat Protocol.

First, create a FastMCP server implementation in a server.py file that exposes the required list_tools and call_tool async methods. Then, in the following agent.py, import the MCP server instance and use it with the MCPServerAdapter:

from uagents import Agent
from uagents_adapter import MCPServerAdapter
from server import mcp

# Create an MCP adapter
mcp_adapter = MCPServerAdapter(
    mcp_server=mcp,
    asi1_api_key="your_asi1_api_key",
    model="asi1-mini"     # Model options: asi1-mini, asi1-extended, asi1-fast
)

# Create a uAgent
agent = Agent()

# Add the MCP adapter protocols to the agent
for protocol in mcp_adapter.protocols:
    agent.include(protocol)

# Run the MCP adapter with the agent
mcp_adapter.run(agent)

Important: When creating MCP tools, always include detailed docstrings using triple quotes (""") to describe what each tool does, when it should be used, and what parameters it expects. These descriptions are critical for ASI:One to understand when and how to use your tools.

For more detailed instructions and advanced configuration options, see the MCP Server Adapter Documentation.

Agentverse Integration

Mailbox Service

By default, agents are created with mailbox=True, which enables the agent to use the Agentverse mailbox service. This allows agents to communicate with other agents without requiring a publicly accessible endpoint.

When mailbox is enabled:

• Agents can be reached by their agent address (e.g., agent1q...)
• No port forwarding or public IP is required
• Messages are securely handled through the Agentverse infrastructure

Agentverse Registration

You can optionally register your agent with the Agentverse API, which makes it discoverable and usable by other users in the Agentverse ecosystem:

1. Obtain an API token from Agentverse.ai
2. Include the token when registering your agent:

   result = register_tool.invoke({
       # ... other parameters
       "api_token": "YOUR_AGENTVERSE_API_TOKEN"
   })
   

When an agent is registered with Agentverse:

• It connects to the mailbox service automatically
• It appears in the Agentverse directory
• A README with input/output models is automatically generated
• The agent gets an "innovationlab" badge
• Other users can discover and interact with it
• You can monitor its usage and performance through the Agentverse dashboard

Example of auto-generated README for LangChain agents:

# Agent Name
Agent Description
![tag:innovationlab](https://img.shields.io/badge/innovationlab-3D8BD3)

**Input Data Model**
```python
class QueryMessage(Model):
    query: str

Output Data Model

class ResponseMessage(Model):
    response: str

Example of auto-generated README for CrewAI agents with parameters:
```markdown
# Agent Name
Agent Description
![tag:innovationlab](https://img.shields.io/badge/innovationlab-3D8BD3)

**Input Data Model**
```python
class ParameterMessage(Model):
    topic: str
    max_results: int | None = None

Output Data Model

class ResponseMessage(Model):
    response: str

Example Query

Research about artificial intelligence

## License

Apache 2.0