Connect your chatbot
Add the management MCP in Claude, ChatGPT, Claude Code, or Gemini CLI and approve the connection.
Every client uses the same address:
https://agentes.productos-ai.com/mcp
The connection uses OAuth: there are no tokens to copy or paste. The first time, the chatbot opens the Agentes console so you can sign in and approve. You also find the URL in the console, under Settings → Chatbots.
Claude (web and Desktop)#
- 1
Open connectors
In Claude, go to Settings → Connectors and choose Add custom connector.
- 2
Paste the URL
Give it a name, such as Productos AI Agentes, and paste the MCP URL. You don't need the advanced settings.
- 3
Connect and approve
Click Connect. The Agentes console opens: sign in, pick the workspace, and approve.
- 4
Turn it on in the chat
In a conversation, turn the connector on from the + button in the message box.
Tip
On Claude's Team and Enterprise plans, an organization owner adds the connector first, and then each member connects it with their own account.
ChatGPT#
- 1
Turn on developer mode
In ChatGPT, go to Settings → Apps & Connectors → Advanced settings and turn on developer mode. Depending on your plan, the menu may be named differently.
- 2
Create the connector
Choose Create, give it a name, and paste the MCP URL. For authentication pick OAuth. ChatGPT registers itself; you don't need a client ID.
- 3
Approve
ChatGPT opens the Agentes console. Sign in, pick the workspace, and approve.
Claude Code#
From your terminal:
claude mcp add --transport http agentes https://agentes.productos-ai.com/mcp
Then, inside Claude Code, run /mcp, choose agentes, and follow the link to approve in the browser.
Gemini CLI#
Add the server to ~/.gemini/settings.json (or your project's .gemini/settings.json):
{
"mcpServers": {
"agentes": {
"httpUrl": "https://agentes.productos-ai.com/mcp"
}
}
}
Then, inside Gemini CLI, run /mcp auth agentes and approve in the browser.
Other clients#
Any client that supports remote MCP over Streamable HTTP with OAuth 2.1 can connect. The server publishes everything a client needs to discover it and register:
| What | Where |
|---|---|
| MCP server | https://agentes.productos-ai.com/mcp |
| Resource metadata | https://agentes.productos-ai.com/.well-known/oauth-protected-resource/mcp |
| Authorization metadata | https://agentes.productos-ai.com/.well-known/oauth-authorization-server |
| Dynamic client registration | https://agentes.productos-ai.com/register |
Authorization uses PKCE with S256.
The approval screen#
When you connect, the console shows:
- Which chatbot is asking for access and which address you'll return to after approving. If you don't recognize the address, cancel.
- The workspaces you administer. Pick one: each connection works in a single workspace. To use several, connect the chatbot once for each.
- What it will be able to do, depending on your plan.
Important
Only admins and owners can approve. If you are a manager or viewer, you'll see a notice: ask an admin to connect the chatbot or to give you that role.
Check that it works#
Ask the chatbot:
"List my Productos AI agents and tell me which one is published."
If it answers with your agents, you're all set. Continue with Create your first agent.