bizzabo-mcp

MCP server for the Bizzabo event management API. Exposes 12 tools for reading events, sessions, speakers, contacts, partners, agenda items, and registrations — wrapped in OAuth2 auth, auto-pagination, 60s caching, and 429 retry-with-backoff.

Quick start

You need three Bizzabo OAuth2 credentials: BIZZABO_CLIENT_ID, BIZZABO_CLIENT_SECRET, BIZZABO_ACCOUNT_ID. Get them from your Bizzabo admin → Integrations → API → OAuth2 Client Credentials. Then pick the client you're using below.

Use with Claude (Desktop or Code)

Drop this into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), or ~/.config/claude/mcp.json for Claude Code:

{
  "mcpServers": {
    "bizzabo": {
      "command": "npx",
      "args": ["-y", "@capitalthought/bizzabo-mcp"],
      "env": {
        "BIZZABO_CLIENT_ID": "...",
        "BIZZABO_CLIENT_SECRET": "...",
        "BIZZABO_ACCOUNT_ID": "..."
      }
    }
  }
}

Restart Claude. Ask "list my Bizzabo events" — it should call list_events.

Use with ChatGPT

MCP works with ChatGPT two different ways depending on what you have:

Option A — ChatGPT Desktop (Developer Mode, Plus/Pro/Team/Enterprise)

OpenAI's Desktop app added local MCP support via Developer Mode.

1. In ChatGPT Desktop → Settings → Connectors → Advanced → enable Developer Mode.
2. Settings → Connectors → Create → choose Local MCP server.
3. Fill in:
   • Name: Bizzabo
   • Command: npx
   • Arguments: -y @capitalthought/bizzabo-mcp
   • Environment variables: add BIZZABO_CLIENT_ID, BIZZABO_CLIENT_SECRET, BIZZABO_ACCOUNT_ID
4. Save, then toggle the connector on in the new-chat composer.
5. Ask "list my upcoming Bizzabo events" — you'll see tool-call cards for list_events etc.

Note: Local MCP connectors only work in the Desktop app (not chatgpt.com in a browser). The set of plans that can enable Developer Mode has shifted over time — if you don't see the setting, check OpenAI's connector docs for your plan.

Option B — OpenAI Agents SDK (for developers)

If you're building on top of OpenAI's API, the Agents SDK can launch the MCP server directly.

Python:

from agents import Agent, Runner
from agents.mcp import MCPServerStdio

async def main():
    async with MCPServerStdio(
        name="bizzabo",
        params={
            "command": "npx",
            "args": ["-y", "@capitalthought/bizzabo-mcp"],
            "env": {
                "BIZZABO_CLIENT_ID": "...",
                "BIZZABO_CLIENT_SECRET": "...",
                "BIZZABO_ACCOUNT_ID": "...",
            },
        },
    ) as bizzabo:
        agent = Agent(
            name="Event Assistant",
            instructions="Use Bizzabo to answer questions about events.",
            mcp_servers=[bizzabo],
        )
        result = await Runner.run(agent, "List the next 5 events.")
        print(result.final_output)

TypeScript (@openai/agents):

import { Agent, run, MCPServerStdio } from "@openai/agents";

const bizzabo = new MCPServerStdio({
  name: "bizzabo",
  command: "npx",
  args: ["-y", "@capitalthought/bizzabo-mcp"],
  env: {
    BIZZABO_CLIENT_ID: "...",
    BIZZABO_CLIENT_SECRET: "...",
    BIZZABO_ACCOUNT_ID: "...",
  },
});

await bizzabo.connect();
const agent = new Agent({
  name: "Event Assistant",
  instructions: "Use Bizzabo to answer questions about events.",
  mcpServers: [bizzabo],
});
console.log((await run(agent, "List the next 5 events.")).finalOutput);
await bizzabo.close();

Use with Gemini

Option A — Gemini CLI (recommended)

Install the CLI if you haven't: npm install -g @google/gemini-cli.

Add to ~/.gemini/settings.json:

{
  "mcpServers": {
    "bizzabo": {
      "command": "npx",
      "args": ["-y", "@capitalthought/bizzabo-mcp"],
      "env": {
        "BIZZABO_CLIENT_ID": "...",
        "BIZZABO_CLIENT_SECRET": "...",
        "BIZZABO_ACCOUNT_ID": "..."
      }
    }
  }
}

Then run gemini and ask "what Bizzabo events are coming up?".

Option B — Gemini API via Python SDK (for developers)

The google-genai SDK auto-translates MCP tool listings into Gemini function-calls:

import asyncio
from google import genai
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client

client = genai.Client()

async def main():
    params = StdioServerParameters(
        command="npx",
        args=["-y", "@capitalthought/bizzabo-mcp"],
        env={
            "BIZZABO_CLIENT_ID": "...",
            "BIZZABO_CLIENT_SECRET": "...",
            "BIZZABO_ACCOUNT_ID": "...",
        },
    )
    async with stdio_client(params) as (read, write):
        async with ClientSession(read, write) as session:
            await session.initialize()
            response = await client.aio.models.generate_content(
                model="gemini-2.0-flash",
                contents="List the next 5 Bizzabo events.",
                config=genai.types.GenerateContentConfig(
                    tools=[session],
                ),
            )
            print(response.text)

asyncio.run(main())

The same pattern works with Google's Agent Development Kit (ADK) — register the stdio server as a toolset in your Agent config.

Use with other MCP clients

Any MCP client that speaks stdio works with the same three pieces:

• Command: npx
• Args: ["-y", "@capitalthought/bizzabo-mcp"]
• Env: BIZZABO_CLIENT_ID, BIZZABO_CLIENT_SECRET, BIZZABO_ACCOUNT_ID

Known-working clients at the time of writing: Claude Desktop, Claude Code, Cursor, Windsurf, Zed, Cline, Continue, OpenAI Agents SDK, Gemini CLI, Google ADK.

Install globally (optional)

If you prefer an installed binary over npx:

npm install -g @capitalthought/bizzabo-mcp
# then use "command": "bizzabo-mcp" in any MCP config above

Environment variables

| Variable | Required | Description | |----------|----------|-------------| | BIZZABO_CLIENT_ID | ✅ | OAuth2 client ID | | BIZZABO_CLIENT_SECRET | ✅ | OAuth2 client secret | | BIZZABO_ACCOUNT_ID | ✅ | Numeric account ID | | DEBUG | — | Set to 1 to log OAuth + request details to stderr |

Tools

| Tool | Description | |------|-------------| | list_events | List all events in the account | | get_event | Get a single event by ID | | list_sessions | List sessions for an event | | get_session | Get a single session by ID | | list_speakers | List speakers for an event | | get_speaker | Get a single speaker by ID | | list_contacts | List contacts in the account | | get_contact | Get a single contact by ID | | list_partners | List partners/sponsors for an event | | get_partner | Get a single partner by ID | | get_agenda | Get an event's agenda | | list_registrations | List registrations for an event |

List endpoints auto-paginate when called without a page argument (up to 10 pages × 100 items = 1,000 records).

Architecture

• OAuth2 client credentials flow against auth.bizzabo.com. Tokens are cached in-memory and refreshed 5 minutes before expiry.
• 60-second response cache (per URL + sorted query params) to avoid hammering the API on repeated identical calls.
• Auto-pagination via getAll() — follows totalPages, inserts 100ms inter-page delay, stops at maxPages and reports truncated: true.
• 429 retry with exponential backoff (1s → 2s, max 2 retries per page).
• Structured errors — auth/rate/server/not-found errors return { error, message } rather than throwing.

Development

npm install
npm run build       # tsc → dist/
npm run dev         # tsc --watch
npm start           # run the server against real Bizzabo

Status

Tests in tests/ are currently stale (written against a pre-OAuth2 API-key design; 12 failing). Fixing them is tracked as a TODO.

License

MIT