AIRA HUB HAS A NEW VERSION THAT WORKS ON CLAUDE OR OTHER MCP CLIENTS THAT USE JSON OR MCP REMOTE.

---> NEW VERSION https://github.com/IhateCreatingUserNames2/AiraHub2/tree/main

Test demo on: https://airahub2.onrender.com/

CLAUDE JSON CONFIG:

{ 
  "mcpServers": { 
    "aira-hub": { 
      "command": "npx", 
      "args": [ 
        "mcp-remote",
        "https://airahub2.onrender.com/mcp/stream"
      ]
    }
  } 
}

AIRA Hub is a FastAPI-based platform for managing MCP (Model Context Protocol) tools and A2A (Agent-to-Agent) skills with OAuth 2.1 authentication. This document provides setup instructions, usage examples, and API reference for the system.

1. Setup Instructions
2. Authentication
3. Agent Connection
4. Tool/Skill Management
5. API Reference
6. Example Clients
7. Troubleshooting

• Python 3.8+
• Required packages:

  pip install fastapi uvicorn httpx pydantic jwt python-jose aiohttp sse-starlette

python AiraHub.py --host 0.0.0.0 --port 8015

The API documentation will be available at http://localhost:8015/docs

The system comes with these pre-configured users:

• admin / password123 (admin privileges)
• agent1 / password123 (agent privileges)

curl -X POST "http://localhost:8015/token" \
  -d "username=admin&password=password123" \
  -H "Content-Type: application/x-www-form-urlencoded"

Response:

{
  "access_token": "eyJhbGciOi...",
  "token_type": "bearer",
  "expires_at": 1234567890
}

Refresh tokens are automatically handled via HTTP-only cookies. To manually refresh:

curl -X POST "http://localhost:8015/refresh-token" \
  -H "Cookie: refresh_token=your_refresh_token"

# Example from dummy_agent_sse.py
async with session.get(
    f"{AIRA_HUB}/connect/stream?agent_url={AGENT_URL}&name={AGENT_NAME}&aira_capabilities={CAPABILITIES}",
    headers={"Authorization": f"Bearer {token}"}
) as resp:
    # Handle SSE events

init_payload = {
    "url": AGENT_URL,
    "description": "Agent description",
    "mcp_tools": [{
        "name": "sample_tool",
        "inputSchema": {"type": "object"}
    }],
    "tags": ["test"]
}

await session.post(f"{AIRA_HUB}/connect/stream/init", json=init_payload)

# From aira_test_client.py
resp = requests.get(f"{AIRA_HUB}/mcp/tools", headers=headers)

tool_request = {
    "agent_url": "http://agent-url",
    "tool_name": "tool_name",
    "arguments": {"param": "value"}
}

resp = requests.post(f"{AIRA_HUB}/mcp/invoke-tool", json=tool_request)

resp = requests.get(f"{AIRA_HUB}/tools/calls/{call_id}")

Two example clients are provided:

1. dummy_agent_sse.py - Demonstrates how to connect an agent via SSE
2. aira_test_client.py - Shows how to interact with the hub as a client

To run them:

python dummy_agent_sse.py
python aira_test_client.py

1. Authentication Failures

   • Verify username/password
   • Check token expiration

2. SSE Connection Drops

   • Ensure your agent maintains the connection
   • Implement reconnection logic

3. Tool Invocation Errors

   • Verify the agent is online
   • Check tool name and parameters match the schema

For debugging, run the server with:

python AiraHub.py --host 0.0.0.0 --port 8015 --log-level debug

This documentation reflects the actual implementation in the provided code files. For the most up-to-date information, always refer to the interactive API docs at /docs when the server is running.