Under 2-minute setup

Bubble Lab MCP Server

Connect Claude Desktop, Claude Code, or Cursor to Bubble Lab in under 2 minutes. Sign up, generate a key, paste one JSON block.

Build workflows across Slack, Salesforce, Google Calendar, Gmail, Notion, HubSpot, Jira, Linear — 49+ integrations.

MCP endpoint: https://api.bubblelab.ai/mcp

Watch MCP demo

April 15, 20261 min
Claude CodeMCPAI Agent

Claude Code + Bubble MCP: Build Production-Ready Automations Fast

Give your Claude Code a standard automation framework. We just launched the Bubble app MCP so you can build reliable automations using your tools without having to wire everything together. In this video, we show you how Claude can directly access your tools through Bubble Lab to build scalable workflows, visualize executions, and share them with your team—all from a single prompt.

What the MCP server does

Bubble Lab’s MCP server exposes 32 platform tools across flow management, credentials, bubbles, capabilities, the Pearl agent, Slack bot building, and tracing. Through these tools your AI client can build and run workflows that touch 49+ apps — Slack, Salesforce, Google Calendar, Gmail, Notion, HubSpot, Jira, Linear, and more.

Example: what you can ask Claude Desktop

  • “Create a flow that, when a new Google Calendar event starts, looks up attendees in Salesforce and posts a prep document to #sales-prep in Slack.”
  • “Every Monday at 9am, summarize last week’s closed deals from HubSpot and post the digest to Slack.”
  • “When a new GitHub issue is labelled p0, create a Linear ticket and @mention the on-call engineer in Slack.”

Flow management

  • list-flows List all flows in your workspace
  • get-flow Fetch a flow’s definition
  • get-flow-bubble Inspect a single bubble in a flow
  • create-flow Create a new flow
  • copy-flow Duplicate an existing flow
  • edit-flow Edit a flow’s bubbles or config
  • activate-flow Turn a flow on
  • delete-flow Delete a flow
  • run-flow Execute a flow synchronously
  • list-executions List past flow executions
  • inspect-execution Read execution logs
  • read-flow-rules Read the flow-authoring rules
  • get-skeleton Scaffold a new flow

Credentials

  • list-credentials List saved credentials
  • create-credential Create a credential
  • initiate-credential-creation Start an OAuth credential flow
  • set-credentials Attach credentials to a flow

Bubbles & registry

  • list-bubbles List all available bubbles
  • get-bubble Get a bubble’s schema

Capabilities (MCP-of-MCPs)

  • list-capabilities List connected capability tools
  • get-capability Get a capability’s details
  • search-capability-tools Fuzzy-search tools across capabilities
  • use-capability-tool Invoke a capability tool

Pearl (agent)

  • create-pearl Create a Pearl agent session
  • reset-pearl Reset a Pearl session
  • pearl-chat-message Send a message to Pearl
  • pearl-list-sessions List Pearl sessions

Slack bot builder

  • create-slack-bot Create a Slack bot
  • list-slack-bots List Slack bots
  • invoke-agent-flow Trigger a flow from a Slack bot

Traces & status

  • list-agent-traces List agent traces for debugging
  • get-status Check server health

Setup

Last updated: May 2026

Step 1 — Create a Bubble Lab account

  1. Go to https://app.bubblelab.ai.
  2. Click “Sign up” and complete registration.
  3. Confirm your email address if prompted.

Step 2 — Generate an MCP API key

  1. Inside Bubble Studio, open Settings → MCP Integrations.
  2. Click Generate New Key, name it (e.g., “Claude Desktop”), and pick an access level:
    • read — list and inspect only
    • write — read + create / edit
    • execute — full access, including running flows
  3. Copy the bl_mcp_ key. It is shown once — store it securely.

Step 3 — Connect your AI client

Replace <your_key_here> with the key from Step 2.

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "bubblelab-cloud": {
      "type": "http",
      "url": "https://api.bubblelab.ai/mcp",
      "headers": {
        "Authorization": "Bearer bl_mcp_<your_key_here>"
      }
    }
  }
}

Claude Code (CLI)

claude mcp add --transport http bubblelab-cloud https://api.bubblelab.ai/mcp \
  --header "Authorization: Bearer bl_mcp_<your_key_here>"

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "bubblelab-cloud": {
      "url": "https://api.bubblelab.ai/mcp",
      "headers": {
        "Authorization": "Bearer bl_mcp_<your_key_here>"
      }
    }
  }
}

Step 4 — Verify the connection

  1. Fully quit and restart your AI client.
  2. Ask it: “List the Bubble Lab tools available to you.” You should see list-flows, create-flow, and the rest.
  3. Try: “Create a Bubble Lab flow that posts a daily standup summary to Slack.”

Troubleshooting

401 Unauthorized
Confirm the header reads Authorization: Bearer bl_mcp_... and the key wasn’t truncated on paste. Also check the key’s access level covers the tool you’re calling.
Tools don’t appear in the client
Fully quit the client and relaunch — Claude Desktop and Cursor only read MCP config at startup.
Wrong endpoint URL
Use https://api.bubblelab.ai/mcp. The server speaks Streamable HTTP, not stdio.

FAQ

How long does setup take?
Under 2 minutes once you have a Bubble Lab account — generate a key, paste one JSON block, restart your client.
What is the Bubble Lab MCP endpoint URL?
https://api.bubblelab.ai/mcp — Streamable HTTP transport. Authenticate with Authorization: Bearer bl_mcp_<your_key>.
Do I need a paid plan to use MCP?
No. MCP access is available on the free Starter plan. Paid plans raise execution limits and active-flow counts.
Which clients are supported?
Claude Desktop, Claude Code (CLI), and Cursor are supported out of the box. Any client that speaks Streamable HTTP MCP with bearer-token auth works.
Which apps can my AI client automate via Bubble Lab?
Slack, Salesforce, Google Calendar, Gmail, Google Sheets, Notion, HubSpot, Jira, Linear, GitHub, Airtable, Stripe, Zendesk, PostgreSQL, Snowflake, and 40+ more. Full list: https://bubblelab.ai/integrations
How do triggers work?
Flows run on demand (invoked by your AI client) or on a cron schedule configured in Bubble Studio.
What do the access levels (read / write / execute) mean?
"read" allows listing and inspecting flows, credentials, executions, and traces. "write" adds creating, editing, activating, and deleting flows, plus managing credentials and creating Pearl sessions / Slack bots. "execute" adds running flows, invoking capability tools, sending Pearl messages, and triggering agent flows. Pick the narrowest scope the client needs.
Is my data safe?
Your API key is hashed at rest. You can revoke any key from Settings → MCP Integrations. Set the access level to "read" if you only want read-only tool calls.
I see a 401 error. What now?
Double-check the key was pasted with the bl_mcp_ prefix, and confirm your access level is sufficient (e.g., "read" cannot run flows).
The tools don’t show up in my client.
Fully quit and restart the client — Claude Desktop and Cursor only load MCP servers at startup.

Ready to connect in under 2 minutes?

Create a Bubble Lab account