MCP (Model Context Protocol)

Endpoint

Server URL: https://mcp.hurozo.com/sse

Auth: Use your Hurozo API token.

MCP Server Config

Build a TOML config for an MCP server. Inputs: server_name (section label), mcp_server (npm package/binary), Add ENV key/values, and Add Arg entries. Output: config_base64.

[mcp_servers.server-name]
command = "npx"
args = ["-y", "mcp-server-input", "--flag=value"]
env = { "KEY1" = "VALUE1", "KEY2" = "VALUE2" }

MCP Server Config (SSE)

Build a TOML config for an SSE-based MCP server. Inputs: server_name, rpc_url, sse_url, and Add ENV key/values (emitted as headers). Output: config_base64.

[mcp_servers.server-name]
transport = "http-sse"
rpc_url = "https://example.com/rpc"
sse_url = "https://example.com/sse"
headers = { "KEY1" = "VALUE1" }
startup_timeout_ms = 35000

Brain Config (moved)

The Brain MCP helper now lives under Hurozo/Brain/BrainConfig. Existing graphs keep using it thanks to automatic name translation.

Read the Brain docs →

MCP Config Merger

Combine multiple config_base64 outputs from MCP Server Config into a single base64 TOML config. Use a multi-connection input to feed many configs; duplicates by server label are de-duplicated with last-write-wins. Output: config_base64.

OpenAI Example

Configure the OpenAI Responses API with an MCP tool to route requests to Hurozo:

from openai import OpenAI

client = OpenAI()

resp = client.responses.create(
    model="gpt-5",
    tools=[
        {
            "type": "mcp",
            "server_label": "hurozo-mcp-server",
            "server_description": "Hurozo agents",
            "authorization": "<your api token>",
            "server_url": "https://mcp.hurozo.com/sse",
            "require_approval": "never",
        },
    ],
    input="What are my upcoming calendar appointments?",
)

Replace <your api token> with a valid Hurozo API token.

What You Get

Discovery of published agents and tools, parameter schemas, and secure execution via SSE. Works with tool‑calling models and MCP‑aware runtimes.