Skip to main content

​
Hosted endpoint

Cyberwave provides a hosted MCP server so teams can connect agents without running local infrastructure.
  • Endpoint: https://mcp.cyberwave.com/mcp
  • Transport: streamable-http
  • Protocol: JSON-RPC 2.0 over MCP
At the moment, the hosted endpoint does not require client-side auth headers. If your organization needs stricter access control, self-host the MCP server behind your own auth layer.

​
Quick setup

​
Cursor

Create .cursor/mcp.json in your project: 
{
  "mcpServers": {
    "cyberwave": {
      "url": "https://mcp.cyberwave.com/mcp"
    }
  }
}
Restart Cursor after editing the file.

​
Claude API (remote MCP connector)

import anthropic

client = anthropic.Anthropic()

response = client.beta.messages.create(
    model="claude-opus-4-1",
    max_tokens=1024,
    messages=[{"role": "user", "content": "List available Cyberwave tools."}],
    mcp_servers=[
        {
            "type": "url",
            "name": "cyberwave",
            "url": "https://mcp.cyberwave.com/mcp"
        }
    ],
    tools=[{"type": "mcp_toolset", "mcp_server_name": "cyberwave"}],
)

​
OpenAI Agents SDK (Python)

from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp

mcp_server = MCPServerStreamableHttp(
    params={"url": "https://mcp.cyberwave.com/mcp"},
    name="cyberwave",
)

agent = Agent(
    name="Cyberwave Assistant",
    instructions="Use Cyberwave MCP tools to inspect before acting.",
    mcp_servers=[mcp_server],
)

result = Runner.run_sync(agent, "List my available workspaces.")
print(result.final_output)

​
Best practices for hosted MCP clients

These practices are aligned with the MCP transport specification and major hosted-MCP client docs:
  1. Use Streamable HTTP endpoints for hosted/serverless deployments.
  2. If you are building a custom client, send both accepted response types in Accept:
    • application/json
    • text/event-stream
  3. Use retries with exponential backoff for transient 5xx and network errors.
  4. Apply sensible tool call timeouts and surface structured errors to users.
  5. Prefer inspect-first workflows for Cyberwave:
    • list/get resources first
    • run action tools with execute=false before execute=true
  6. Track observability signals:
    • tool name
    • duration
    • status/error code
    • endpoint latency (p95/p99)

​
Validation and benchmark resources

  • MCP Inspector for manual interoperability testing and tool/resource validation.
  • MCP-Bench for evaluating agent tool-use quality across MCP servers.
  • Cloud Run metrics / your APM for request rate, latency, error rate, and saturation in production.

​
References