Skip to content
BuildBase in your AI tools

Connect once, then ask.

BuildBase in your AI tools: connect Claude, ChatGPT, Cursor or another AI tool to BuildBase's own MCP server, and manage your BuildBase organization from the chat. You sign in as yourself, and the agent can do only what your role and your organization allow.

One address

The same address for every AI tool that connects to remote MCP servers with sign-in.

https://api.console.buildbase.app/mcp
  1. Add the address

    In your AI tool, add the address above as a remote MCP server. Chat apps usually call it a custom connector; coding tools call it an MCP server. Pick the HTTP kind.

  2. Sign in as yourself

    Your tool opens BuildBase's sign-in page. Sign in the way you sign in to the console, then read the consent screen, which names the tool and what it asks for, and approve it.

  3. Ask

    The agent now has BuildBase's 19 tools: it can list your organizations and work in each one that lets agents in, with your role. It can read, and make changes where your organization allows them; it asks you again before an export or a delete.

Connect your tool

Each opens the sign-in page the first time it connects. The setup guide has the steps for each app.

~/.cursor/mcp.jsonJSON
{
  "mcpServers": {
    "buildbase-console": {
      "url": "https://api.console.buildbase.app/mcp"
    }
  }
}
.vscode/mcp.jsonJSON
{
  "servers": {
    "buildbase-console": {
      "type": "http",
      "url": "https://api.console.buildbase.app/mcp"
    }
  }
}

Claude Code

terminalBash
claude mcp add --transport http buildbase-console https://api.console.buildbase.app/mcp

# then run /mcp in Claude Code to sign in

Codex

terminalBash
codex mcp add buildbase-console --url https://api.console.buildbase.app/mcp
codex mcp login buildbase-console

In Claude, add the address as a custom connector and, under OAuth client, choose Register automatically: the default, Claude's published identity, does not work with BuildBase's server. In ChatGPT, add it under Plugins as a custom MCP server. A job with nobody to sign in, such as CI, can send one of your organization's API keys as the bearer instead; the setup guide shows how.

Step by step, for each AI tool

Then ask

The agent reads what you can read in the console, and changes things where your organization allows it. A few things to try:

  • “Which of my organizations is on a trial, and when does it end?”
  • “List the members of Acme and their roles.”
  • “Which workflows does Acme have, and which of them are published?”
  • “Export our audience of contacts created since the first of the month.”
  • “Write a blueprint that adds a beta feature flag and a feedback form, and show me the plan.”
  • “Which environment variables does my Next.js app need?”

What stays with you

The agent works as you, inside rules you and your organization set.

Each organization decides

Every organization has one setting for AI agents: Off, Read only or Read and change. Read only is the default. Owners and admins change it in the console, under Account, then Security, and it applies to every agent at once.

You approve what matters

Deleting something, changing members, roles or webhooks, sending to an audience and moving money each need your approval in the chat, for that exact change. Sign-in settings stay with you in the console.

Secrets never reach the agent

API keys, signing keys, payment and mail credentials and webhook secrets are never sent to the agent, even one it just created. Any secret in what it reads is replaced by a note.

It acts as you, never more

The agent signs in as you and works with your role, your plan and the environments you may open. An owner or admin can block one agent for one member, at once.

Every change is on the record

Each change an agent makes is in the organization's audit log as an AI agent change, naming the tool, above your name.

Read first, more when needed

An agent first asks to read. When a task needs more, such as a change or an export, your tool asks you to approve just that, at the moment it matters.

Frequently Asked Questions

What is the BuildBase MCP server address?

https://api.console.buildbase.app/mcp. Add it to your AI tool as a remote MCP server over HTTP. Your tool opens BuildBase's sign-in page the first time it connects.

Can the agent change things in my organization?

Only where the organization allows it. Each organization is Off, Read only (the default) or Read and change, set by an owner or admin. Even then, deletes, member and role changes, sends and money moves need your approval in the chat.

Does the agent see my API keys or other secrets?

No. Secrets never reach the agent, even one it creates: a webhook it adds comes back without its signing secret. To see a secret, rotate it in the console.

What if I do not have an account yet?

A coding agent can create an organization for you with one request and work in it right away. You keep it by claiming it with the link and code the agent gives you; unclaimed, it is deleted after 72 hours. The agent's instructions are at https://www.buildbase.app/auth.md.

Is this how my own users connect agents to my app?

No, that is Agent-ready apps. This page is about managing BuildBase itself from your AI tool. To give the app you build an MCP server of its own, see https://www.buildbase.app/ai-agents.

Bring BuildBase into your chat

Copy the address, add it to your AI tool and sign in. Your organization stays read only until an owner or admin allows more.

Free plan7-day free trial on paid plansNo credit card required