Getting started
A short walkthrough that takes you from sign-up to your first MCP shared with your team.
You will:
- Create your organization
- Add your first upstream MCP
- Connect your own AI client to the gateway
- Invite a teammate
If any term in here is unfamiliar, see Concepts.
1. Create your organization
Sign in to MCP Hero with Google. The first time you sign in, MCP Hero asks for an organization name — that's the container for your team, your MCPs, and your roles. The short name (used in URLs) auto-fills from the display name; both are editable.
Click Create organization. You will land on the Upstream MCPs page, ready to add your first one.
2. Add your first upstream MCP
Click Add MCP. The wizard has two steps.
Step 1 — Configuration. Pick how to connect:
- URL — paste the URL of an HTTP MCP (Mixpanel, HubSpot, Stripe, and so on).
- JSON — paste a standard
{"mcpServers": {…}}config block. Useful if you already have one from another tool or a teammate, or for stdio MCPs (which are commands, not URLs).
When you paste JSON that uses ${NAME} references, a Variables panel appears so you can fill those values inline. If MCP Hero spots a raw credential in the JSON, it offers to extract it into a Variable for you. Click Next.
Step 2 — Details. Pick an ID and Display Name (auto-filled from the URL or package). For HTTP MCPs, pick the auth mode the upstream uses (the picker is hidden for stdio — see below). For stdio MCPs, pick the sandbox Resources (CPU / memory / disk).
Which authentication mode?
See User authentication modes in Concepts. Short version: pick None if the MCP doesn't need user-level auth (it's open, or you'll paste a shared API key into its server config), Shared if it's a workspace you'll sign into once on the team's behalf, or Per-user if every team member should sign in for themselves. Stdio MCPs are always None — they authenticate via Variables and Files instead (see Upstream MCPs).
Click Add. MCP Hero connects to the upstream and discovers its tools. The new MCP appears on the Upstream MCPs page.
Detailed flows for each connection style — including stdio MCPs and OAuth-protected MCPs — are in Upstream MCPs. For credential-handling specifics on stdio MCPs, see Stdio MCP authentication.
3. Connect your own AI client to the gateway
Open Connect AI Assistant in the sidebar. Your gateway URL ends in /mcp — that single URL is all your AI client needs. Through it, your client will see every upstream MCP you have access to.
Pick your client from the tabs and follow the on-page steps. MCP Hero ships built-in instructions for:
- Claude.ai
- Claude Desktop
- Claude Code
- ChatGPT
- Cursor
My client is not listed
Any MCP client that speaks the Streamable HTTP transport will work. Use the gateway URL directly, or copy the JSON config from the Advanced: JSON configuration disclosure.
Once connected, ask your AI client to list its available tools — you should see your new upstream's tools, prefixed with the upstream name (for example mixpanel__query_events).
4. Invite a teammate
Open the Team page.
- Click Add Member, enter their Google email, pick a role (it defaults to user), and save.
- Copy the invite link at the top of the page and send it to them.
When they click the link and sign in with Google, they land directly in your organization with the role you assigned.
Members have to be added before they can join. Someone who signs in to MCP Hero with Google but hasn't been added to your team won't see your organization — they'll be prompted to create their own.
You now have an organization with one MCP, one teammate, and one AI client connected.
Next steps:
- Add more upstream MCPs — see Upstream MCPs.
- Set up roles for different access levels — see Roles and permissions.
- Manage MCP Hero conversationally from your AI client — see Admin MCP.