Home / Araçlar / MCP & Connectors

Built-in Tool

MCP & Connectors Beta

Connect GlomaxGPT models to any external service through the Model Context Protocol (MCP). Use GlomaxGPT-maintained connectors for popular SaaS tools, or build and host your own MCP server to expose custom tools to your agents.

What is the Model Context Protocol?

The Model Context Protocol (MCP) is an open standard that defines how AI models communicate with external tool servers. An MCP server exposes a set of named tools with typed inputs and outputs. The model can discover these tools, call them, and process the results — all transparently over a standardized JSON-RPC protocol.

Open Standard

MCP is an open protocol supported by multiple AI providers. Build once, connect to any compatible model or platform.

Extensible

Any service with an HTTP API can be wrapped as an MCP server. Expose database queries, internal APIs, or third-party SaaS tools as model-callable functions.

Secure by Default

MCP servers authenticate via OAuth 2.0, API keys, or mTLS. GlomaxGPT manages credential exchange so secrets never flow through the model context.

Remote MCP Only — the GlomaxGPT Responses API connects to remote MCP servers over HTTPS. Local (stdio) MCP servers are not supported directly; wrap them in an HTTP server to use them with the API.

Connecting a Remote MCP Server

Add a mcp tool entry to your tools array with the server URL and any required headers. The model will automatically discover available tools from the server's manifest and invoke them as needed.

Python — Connect Remote MCP Server
from GlomaxGPT import GlomaxGPT

client = GlomaxGPT()

response = client.responses.create(
    model="glomaxgpt-ultra",
    input="List the open GitHub issues in the GlomaxGPT/GlomaxGPT-python repository.",
    tools=[{
        "type": "mcp",
        "server_label": "github",
        "server_url": "https://mcp.github.com/v1",
        "headers": {
            "Authorization": f"Bearer {GITHUB_TOKEN}"
        }
    }]
)

print(response.output_text)

Tool Approval

By default, the model can call any tool on the MCP server. You can restrict which tools are allowed using allowed_tools, or require human approval before any tool is invoked using require_approval.

Python — With Tool Restrictions
response = client.responses.create(
    model="glomaxgpt-ultra",
    input="Summarize the latest Slack messages in #engineering",
    tools=[{
        "type": "mcp",
        "server_label": "slack",
        "server_url": "https://mcp.slack.com/v1",
        "headers": {"Authorization": f"Bearer {SLACK_TOKEN}"},
        "allowed_tools": ["list_channels", "get_messages"],
        "require_approval": "never"  # or "always" for human review
    }]
)

GlomaxGPT-Maintained Connectors

GlomaxGPT maintains a set of first-party MCP connectors for popular services. These are production-ready, regularly updated, and designed with security best practices built in.

Slack

Available

Read channels, post messages, search message history, manage reactions, and interact with workspace members.

list_channels get_messages post_message search

Server: https://mcp.slack.com/v1

GitHub

Available

Manage repositories, issues, pull requests, code search, and GitHub Actions workflows.

list_issues create_pr search_code get_file

Server: https://mcp.github.com/v1

Gmail

Available

Read, search, compose, and send emails. Manage labels, threads, and Gmail filters programmatically.

search_emails send_email get_thread label

Server: https://mcp.gmail.GlomaxGPT.com/v1

Salesforce

Available

Query and update CRM records, leads, opportunities, accounts, and run SOQL queries against your Salesforce org.

soql_query create_record update_record search

Server: https://mcp.salesforce.GlomaxGPT.com/v1

Google Calendar

Preview

Read and create calendar events, check availability, schedule meetings, and manage calendar invitations.

list_events create_event check_availability

Server: https://mcp.gcal.GlomaxGPT.com/v1

Notion

Preview

Read and write Notion pages and databases. Search workspace content, create documents, and update properties.

search get_page create_page query_db

Server: https://mcp.notion.GlomaxGPT.com/v1

Stripe

Preview

Query payments, customers, subscriptions, invoices, and products in your Stripe account. Read-only by default.

list_customers get_payment list_invoices

Server: https://mcp.stripe.GlomaxGPT.com/v1

Linear

Coming Soon

Manage Linear issues, projects, and cycles. Create tasks, update statuses, and query your engineering roadmap.

list_issues create_issue update_status

Coming Q3 2026

Kimlik Doğrulama

MCP servers can authenticate using several methods. Choose the approach that best fits your server's security requirements.

Python — Bearer Token
tools=[{
    "type": "mcp",
    "server_label": "my_service",
    "server_url": "https://api.myservice.com/mcp",
    "headers": {
        "Authorization": f"Bearer {MY_ACCESS_TOKEN}"
    }
}]
Python — OAuth 2.0
# For OAuth 2.0, obtain the access token via your OAuth flow
# then pass it as a Bearer token in the Authorization header
import requests

token_response = requests.post("https://oauth.myservice.com/token", data={
    "grant_type": "client_credentials",
    "client_id": CLIENT_ID,
    "client_secret": CLIENT_SECRET,
    "scope": "mcp:read mcp:write"
})
access_token = token_response.json()["access_token"]

tools = [{
    "type": "mcp",
    "server_label": "my_service",
    "server_url": "https://api.myservice.com/mcp",
    "headers": {"Authorization": f"Bearer {access_token}"}
}]
Python — API Key Header
tools=[{
    "type": "mcp",
    "server_label": "my_service",
    "server_url": "https://api.myservice.com/mcp",
    "headers": {
        "X-API-Key": MY_API_KEY,
        "X-Workspace-ID": MY_WORKSPACE_ID
    }
}]

Build Your Own MCP Server

Any web server that implements the MCP specification can serve as a tool server. The server must expose a /manifest endpoint listing available tools and a /call endpoint that executes them.

Python — Minimal MCP Server (FastAPI)
from fastapi import FastAPI, Request
from pydantic import BaseModel
import httpx

app = FastAPI()

# Tool manifest — lists all tools this server exposes
MANIFEST = {
    "schema_version": "v1",
    "name": "weather_server",
    "description": "Provides real-time weather information",
    "tools": [{
        "name": "get_weather",
        "description": "Returns current weather for a given city",
        "input_schema": {
            "type": "object",
            "properties": {
                "city": {"type": "string", "description": "City name"},
                "units": {"type": "string", "enum": ["celsius", "fahrenheit"]}
            },
            "required": ["city"]
        }
    }]
}

@app.get("/manifest")
async def get_manifest():
    return MANIFEST

class CallRequest(BaseModel):
    tool: str
    input: dict

@app.post("/call")
async def call_tool(req: CallRequest):
    if req.tool == "get_weather":
        city = req.input["city"]
        units = req.input.get("units", "celsius")
        # Call your actual weather API here
        return {
            "result": {
                "city": city,
                "temperature": 22,
                "units": units,
                "condition": "Partly cloudy"
            }
        }
    return {"error": f"Unknown tool: {req.tool}"}

Full Python Example

Combining multiple MCP connectors in a single agent that can read GitHub issues and post a Slack summary.

Python — Multi-Connector Agent
import os
from GlomaxGPT import GlomaxGPT

client = GlomaxGPT()

response = client.responses.create(
    model="glomaxgpt-ultra",
    input="Fetch all open issues labeled 'bug' from GlomaxGPT/GlomaxGPT-python, summarize them, and post the summary to the #bugs channel in Slack.",
    tools=[
        {
            "type": "mcp",
            "server_label": "github",
            "server_url": "https://mcp.github.com/v1",
            "headers": {"Authorization": f"Bearer {os.environ['GITHUB_TOKEN']}"},
            "allowed_tools": ["list_issues", "get_issue"]
        },
        {
            "type": "mcp",
            "server_label": "slack",
            "server_url": "https://mcp.slack.com/v1",
            "headers": {"Authorization": f"Bearer {os.environ['SLACK_BOT_TOKEN']}"},
            "allowed_tools": ["post_message"],
            "require_approval": "always"  # Confirm before posting
        }
    ],
    instructions="You are a DevOps assistant. Be concise and structured in your summaries."
)

print(response.output_text)
MCP Ecosystem — Beyond GlomaxGPT connectors, thousands of community-built MCP servers are available. Browse the MCP registry at mcp.so for databases, APIs, internal tools, and more.

Sonraki Adımlar

Use MCP connectors together with Agents to build powerful multi-step workflows that span multiple services and data sources.

Agents Guide → All Tools →