Connect BankSync with the Python SDK

Build your own MCP client using the official mcp Python package.

Prerequisites

A BankSync account with at least one connected bank
An API key with all permission scopes (created from Workspace menu → Developers)

Install the SDK

Terminal

pip install mcp

Connect to BankSync

Custom MCP Client

1import asyncio
2from mcp import ClientSession
3from mcp.client.streamable_http import streamablehttp_client
4
5async def main():
6    async with streamablehttp_client(
7        "https://mcp.banksync.io",
8        headers={"X-API-Key": "bsk_your_api_key_here"},
9    ) as (read_stream, write_stream, _):
10        async with ClientSession(read_stream, write_stream) as session:
11            await session.initialize()
12
13            # List available tools
14            tools = await session.list_tools()
15            print(f"Connected! {len(tools.tools)} tools available")
16
17            # Call a tool
18            result = await session.call_tool(
19                "list_workspaces",
20                arguments={},
21            )
22            print(result)
23
24asyncio.run(main())

Verify Your Connection

After configuring your client, restart it and try these prompts to verify the connection is working:

“List my workspaces”

“What banks do I have connected?”

“What’s my checking account balance?”

Troubleshooting

“No tools available” or MCP server not listed

Make sure you restarted your AI client after editing the configuration file.
Verify the JSON is valid (no trailing commas, correct brackets).
Check that the URL is exactly https://mcp.banksync.io.

“Authentication failed” or “Invalid API key”

Confirm the key starts with bsk_ and has not been revoked.
Ensure the key is in the headers object under X-API-Key.
Check for accidental whitespace or newlines around the key value.

Tools are listed but return errors

Verify your API key has the required scopes. For full access, select all scopes when creating the key.
Make sure you have at least one bank connected in the BankSync app.
Check that you are passing the correct workspace ID when prompted.

Connection timeout or network errors

The MCP server uses Streamable HTTP transport. Ensure your network or proxy allows outbound HTTPS connections.
If using a VPN or firewall, allow connections to mcp.banksync.io.
Try the connection from a different network to rule out local issues.
If your client does not support remote HTTP servers, use the npm package instead: npx -y @banksync/mcp

Debugging with MCP Inspector

Use the MCP Inspector to test the connection independently of your AI client.
Run: npx @modelcontextprotocol/inspector --url https://mcp.banksync.io --header "X-API-Key: bsk_your_key"
If the Inspector connects successfully but your client does not, the issue is with your client configuration.