Create Your First Agent

This guide walks you through creating and using your first MCP agent.

Prerequisites

• GateFlow account with API key
• At least one provider configured (OpenAI, Anthropic, etc.)

Step 1: Create an Agent

curl -X POST https://api.gateflow.ai/v1/mcp/agents \
  -H "Authorization: Bearer gw_prod_..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "My First Agent",
    "description": "A simple chatbot agent",
    "permissions": {
      "tools": ["llm/chat"],
      "models": ["gpt-4o-mini"]
    }
  }'

Response:

{
  "agent_id": "agent_abc123",
  "name": "My First Agent",
  "api_key": "gf-agent-xyz789...",
  "mcp_endpoint": "https://mcp.gateflow.ai/agent_abc123",
  "permissions": {
    "tools": ["llm/chat"],
    "models": ["gpt-4o-mini"]
  }
}

Save Your Agent Key

The api_key is only shown once. Store it securely.

Step 2: Connect to the Agent

Using Python

import httpx

class MCPAgent:
    def __init__(self, endpoint: str, api_key: str):
        self.endpoint = endpoint
        self.api_key = api_key
        self.client = httpx.Client()

    def call_tool(self, tool: str, params: dict):
        response = self.client.post(
            f"{self.endpoint}/tools/{tool}",
            headers={"Authorization": f"Bearer {self.api_key}"},
            json=params
        )
        return response.json()

agent = MCPAgent(
    endpoint="https://mcp.gateflow.ai/agent_abc123",
    api_key="gf-agent-xyz789..."
)

Using LangChain

from langchain_community.tools import MCPToolkit

toolkit = MCPToolkit(
    server_url="https://mcp.gateflow.ai/agent_abc123",
    api_key="gf-agent-xyz789..."
)

tools = toolkit.get_tools()

Step 3: Use a Tool

# Call the chat tool
result = agent.call_tool("llm/chat", {
    "model": "gpt-4o-mini",
    "messages": [
        {"role": "user", "content": "Hello! What can you do?"}
    ]
})

print(result["response"]["content"])

Output:

Hello! I'm an AI assistant. I can help you with:
- Answering questions
- Having conversations
- Providing information

How can I help you today?

Step 4: Check Usage

# Use the self-inspect tool
usage = agent.call_tool("self/whoami", {})

print(f"Agent: {usage['agent_name']}")
print(f"Session cost: ${usage['session_cost']}")
print(f"Tools available: {usage['tools']}")

Step 5: View Audit Log

Check what your agent did:

curl https://api.gateflow.ai/v1/mcp/agents/agent_abc123/audit-log \
  -H "Authorization: Bearer gw_prod_..." \
  -G -d "limit=10"

Complete Example

import httpx
import json

# Configuration
AGENT_ENDPOINT = "https://mcp.gateflow.ai/agent_abc123"
AGENT_KEY = "gf-agent-xyz789..."

def chat(message: str) -> str:
    """Send a message to the agent and get a response."""
    response = httpx.post(
        f"{AGENT_ENDPOINT}/tools/llm/chat",
        headers={"Authorization": f"Bearer {AGENT_KEY}"},
        json={
            "model": "gpt-4o-mini",
            "messages": [{"role": "user", "content": message}]
        }
    )
    result = response.json()
    return result["response"]["content"]

# Have a conversation
print(chat("What is the capital of France?"))
print(chat("What's the population?"))
print(chat("Tell me an interesting fact about it."))

Next Steps

• Agent Templates - Pre-built agent configurations
• Connecting Clients - More client options
• First Tool Invocation - Deep dive into tools