Connect the o365hq.com MCP Server to ChatGPT, Claude, Copilot, Gemini and Grok: Microsoft 365 Prices, Quotes and Services from Your Own AI Agent

MCP, the Model Context Protocol, is the open standard that lets an AI assistant call tools on a remote server. The o365hq.com server speaks it over plain HTTPS at https://o365hq.com/mcp: no sign-in, no API key, and it works the same from every client that supports remote MCP servers. Prices come from the current month's US Microsoft CSP price list, which Microsoft changes monthly, and every answer says which month it priced against. The server never buys anything. Quote pages it produces are marked as AI-prepared, and a person at IT Partner confirms every order before anything is provisioned.

If you prefer to read the machine's own description first: the agent guide lists the tools, the changelog lists what changed, and https://o365hq.com/openapi.json describes the same functions for code that does not speak MCP.

Anthropic's support article "Get started with custom connectors using remote MCP", checked 22 October 2026, gives the path for individual plans:

1. Open Customize, then Connectors.
2. Click + Add, then Add custom connector.
3. Name it, for example IT Partner (o365hq), and paste https://o365hq.com/mcp. Click Continue.
4. Claude detects the authentication settings; the server has none, so choose No sign in and click Add.

Custom connectors are available on Free (one connector), Pro, Max, Team and Enterprise. On Team and Enterprise plans an Owner adds it under Organization settings → Connectors → Add, hovers Custom, selects Web, and members then enable it for themselves. Claude Desktop uses the same connector list. In a chat, the tools appear under the connector's name; ask a question and Claude calls them as needed.

Claude Code takes one command:

claude mcp add --transport http o365hq https://o365hq.com/mcp

OpenAI's help article "Developer mode and MCP apps in ChatGPT (beta)" is the reference; the menu labels moved more than once during 2026, so if a step does not match, search Settings for Developer mode.

1. Open Settings, then Apps & Connectors (older builds: Connectors), then Advanced settings, and switch Developer mode on.
2. Back on the connectors page choose Create (or Add custom connector), give it a name and a short description, paste https://o365hq.com/mcp, and pick No authentication. Tick the acknowledgement and create it.
3. In a new chat open the + menu, choose Developer mode, and tick the connector. Ask away.

Developer mode is available on Plus, Pro, Business, Enterprise and Edu plans on the web; Free accounts cannot add custom connectors. On Business and Enterprise workspaces an admin first allows it under Workspace settings → Permissions & Roles → Connected Data → Create custom MCP connectors, and admins cannot switch it on for one member at a time. The feature carries a beta label.

Microsoft 365 Copilot does not take a custom MCP URL from an end user's settings page; custom tools reach it through an agent built in Copilot Studio. Microsoft Learn's "Connect your agent to an existing Model Context Protocol (MCP) server" gives the wizard path:

1. In Copilot Studio open your agent and its Tools page.
2. Choose Add a tool, then New tool, then Model Context Protocol.
3. Enter the server URL https://o365hq.com/mcp, transport Streamable HTTP (Copilot Studio stopped supporting SSE in August 2025), authentication None, and finish the wizard.
4. Publish the agent to Microsoft 365 Copilot or Teams. People then ask the agent, and the agent calls the IT Partner tools.

Agents that run inside Microsoft 365 Copilot need the appropriate Copilot Studio and Microsoft 365 Copilot licensing for the people who use them; check Microsoft's current licensing page before rolling one out.

GitHub Copilot in Visual Studio Code reads a workspace file .vscode/mcp.json:

{ "servers": { "o365hq": { "type": "http", "url": "https://o365hq.com/mcp" } } }

Open Copilot Chat in agent mode and the tools are listed under the server name.

Gemini CLI, documented at "MCP servers with the Gemini CLI", takes either a command or a settings entry. The field matters: httpUrl selects the streamable HTTP transport this server uses; url would select the older SSE transport and fail.

gemini mcp add --transport http o365hq https://o365hq.com/mcp

or in ~/.gemini/settings.json:

{ "mcpServers": { "o365hq": { "httpUrl": "https://o365hq.com/mcp" } } }

In the consumer Gemini app, custom servers are added under Connected Apps on the web with the server's HTTPS URL; as of October 2026 Google offers this to accounts with Gemini Spark access and the connected app then shows up in the mobile app too. Gemini Enterprise administrators add a custom MCP server as a data source, which an organization policy blocks by default until an admin allows it. Google's rollout differs by account type and country, so treat the app and Enterprise paths as "check your settings page" rather than a promise.

On grok.com open Connectors, choose New Connector, then Custom, paste https://o365hq.com/mcp, and save; Grok discovers the tools on its own. xAI's plan pages say which tiers include custom connectors, and that has changed during the year, so check before relying on it.

Developers using the xAI API attach the same server as a remote MCP tool, documented under "Remote MCP Tools": a tool of type mcp with server_url set to https://o365hq.com/mcp and a server_label such as o365hq. Streamable HTTP and SSE are the supported transports; this server speaks the first.

Every answer carries a small _meta.o365hq block with the price-list month and the caller's remaining allowances, so a careful agent can cite freshness and pace itself.

Every output below was produced by the live server on 22 October 2026 against the October price list, trimmed for length.

"What does Microsoft 365 Business Premium cost for 10 people?"

The agent calls get_plan_pricing and relays: $22 per user per month on an annual commitment paid annually ($264 per user per year, $2,640 per year for ten), $23.10 per user per month on an annual commitment paid monthly, and $26.40 per user per month month-to-month; last price sync 1 October 2026; Microsoft's US list price, not a binding quote.

"Business Premium or E3 for a 40-person firm?"

compare_plans returns IT Partner's verdict: under 300 users, Business Premium is usually the better-value choice, it held its price in the 2026 update, includes Defender for Business, and its mailboxes now match E3 at 100 GB; choose E3 for Windows 11 Enterprise, unlimited email archiving, full Data Loss Prevention and Standard eDiscovery, or when you are at or growing past the 300-seat ceiling. A fifteen-row feature matrix follows, with the page link.

"We have one host with two 8-core processors, four VMs and 25 users. What Windows Server licensing do we need?"

calculate_core_licensing answers: 32 core licenses as 16 two-core packs, because Standard covers two VMs per fully licensed host and four VMs mean licensing the host twice; 25 Windows Server user CALs; Standard at $2,352 for the packs plus $1,250 for the CALs, $3,602 one-time, against Datacenter at $8,018; Standard is cheaper; and two ready order links plus the calculator page with the inputs filled in.

"Put 10 Business Standard licenses and the GoDaddy Microsoft 365 defederation service on one quote I can send to my boss."

quote_link prices the subscription line (Microsoft 365 Business Standard, 10 users, 1-year commitment paid annually, $168 per user per year, $1,680 per year) and the service line (GoDaddy Microsoft 365 Defederation & Tenant Takeover, $525 per project, one-time), totals of $1,680 due annually and $525 one-time, and returns a shareable live-priced quote page, a print version and a full view. The page is live-priced and nothing is stored; a person confirms before any order.

"Prepare a formal quotation for 25 Microsoft 365 E3 for Example Co."

create_quote with dry_run first: Microsoft 365 E3, 25 users at $39.00 per user per month, $11,700 per year, October price list, with the disclaimer the agent must relay: prepared automatically at your request; check with the IT Partner sales team before purchasing. The committed call then returns a quote number and a /q link that stays valid for 14 days.

"Are we eligible for nonprofit pricing?"

check_eligibility for Charity: nonprofits with recognized charitable status, in the US typically 501(c)(3) organizations plus churches and some others; Microsoft validates eligibility during nonprofit-tenant registration; discounts are significant; IT Partner runs the validation with you at no charge and re-runs any quote in the Charity segment once approved.

"Something is missing."

When an agent needs what the server does not offer, it calls request_feature with the gap and the call it tried. The answer is short: { "ok": true, "id": 12, "status": "new", "repeated": false, "message": "Recorded. A human at IT Partner reads every request the same business day; what ships is announced in the changelog." }.

Anonymous callers get about 60 requests a minute, five human-answered build_quote requests an hour, three committed create_quote pages a day (dry runs are free) and five feature requests a day. A registered-agent key, free for accountable callers, lifts that to about 300 requests a minute and ten committed quotes a day, keys allowances to the agent's name instead of its address, and puts that name on the notification our team receives. Ask for one on the contact page with a line on how you will use it, then send it as an x-agent-key header.

What we ask in return is in the agent guide: relay quote disclaimers verbatim, never present an AI-prepared quote as human-verified, prefer the markdown mirrors over scraping HTML, and remember there is no automated purchasing.

The server grew out of agents asking good questions, and the new request_feature tool is the direct line. If your agent reports that it could not do something, tell it to file the request: what it tried, the tool and arguments, and why your task needed it. The request lands in IT Partner's triage queue, a person reads it the same business day, and shipped changes appear in the changelog with their date. The services-on-a-quote behavior described above started as exactly that kind of report from an agent connected through Claude.

Does connecting cost anything, and is there a sign-in?

No and no. The server is public and free, and prices are Microsoft's published list for the US market. IT Partner is paid by Microsoft as a CSP partner when you buy through it, which is why the prices match buying direct.

Can my agent buy licenses through it?

No. It can price, compare, calculate and prepare quote pages. Every order is confirmed by a person at IT Partner, by design.

What does the server store?

quote_link pages are live-priced from their URL and store nothing. create_quote stores the quotation page it creates. build_quote and request_feature store what you sent so a person can act on it. When your agent passes customer_email, one confirmation message goes to that address and the quote is sent only after the person presses the button in it.

My client is not on the list.

Any client that supports remote MCP servers over streamable HTTP works with the same URL. For code without MCP, https://o365hq.com/openapi.json describes the same functions, and https://o365hq.com/llms.txt maps the site for a model that only reads pages.

How current are the prices?

Each answer names the price-list month it used. Microsoft publishes a new CSP price list monthly and IT Partner syncs it at the start of the month.