Splitwise MCP Server
Enables conversational control of Splitwise accounts through Claude AI, allowing users to add expenses, check group balances, record settlements, and manage payment splits using natural language commands. Supports multiple currencies and flexible splitting methods including equal, exact, and percentage-based divisions.
README
Splitwise MCP Server — Control Splitwise With Plain English
"Add a ₩45,000 dinner, I paid, split equally." Done. No app. No tapping. No math.
Why I Built This
I was about to head on a trip with friends and already dreading the expense tracking.
You know how it goes — someone pays for dinner, someone else grabs the cab, another person covers the hotel deposit. By day two, you've lost track. So you open Splitwise, tap through four screens, type the amount, select the people, pick who paid, hit save. Then do it again for the next thing. And the next.
I thought: I'm already using Claude for everything else. What if I could just tell it what happened, and it handles Splitwise for me?
So I built this over a weekend. It's a small Python server that connects Claude AI directly to your Splitwise account. You talk to Claude like a human, and it does the Splitwise work for you — adding expenses, checking balances, recording settlements, fixing mistakes.
No switching apps. No manual entry. Just say what happened.
What It Feels Like to Use
You: "We had a group dinner for ₩65,000. I paid. Split equally."
Claude: Done! Added "Group dinner" for ₩65,000.00 (KRW), split equally
among 4 members. Each person owes ₩16,250.
You: "What does everyone owe in the trip group?"
Claude: Here are the current balances:
• Raj owes you ₩32,500
• Priya owes you ₩16,250
• Sam is owed ₩8,000 by the group
You: "Record that Raj paid me back ₩30,000."
Claude: Payment of ₩30,000 recorded. Raj's balance updated.
You: "Wait, that dinner was actually ₩72,000. Can you fix it?"
Claude: Updated! "Group dinner" changed to ₩72,000. Shares recalculated.
Works with any currency Splitwise supports — USD, EUR, INR, KRW, JPY, and more.
How It Works (The Simple Version)
You talk to Claude
↓
Claude understands what you want
↓
Claude calls this MCP server
↓
This server calls the Splitwise API
↓
Your Splitwise account updates instantly
↓
Everyone's app reflects the change
MCP (Model Context Protocol) is a standard way to give Claude access to external tools. This project is the "Splitwise tool" — a bridge between Claude and your Splitwise account.
See architecture.md for a detailed diagram of how everything connects.
What You Can Do
| Ask Claude | What Happens |
|---|---|
| "List my groups" | Shows all your Splitwise groups with members |
| "Who owes what in the [group]?" | Shows all balances |
| "Add a $52 lunch, I paid, split equally" | Creates the expense |
| "Add $240 hotel — I paid $120, Alex $80, Sam $40" | Exact split |
| "Split the $35 cab 50/30/20" | Percentage split |
| "Record that Alex paid me back $20" | Settlement payment |
| "Delete that last expense" | Removes it |
| "Change the dinner to $60" | Updates and recalculates |
Setup
What You'll Need
- Python 3.10+ (
python3 --versionto check) - A Splitwise account + API key from secure.splitwise.com/apps
- Claude Desktop or Claude Code
Step 1: Clone and install
git clone https://github.com/yourusername/splitwise-mcp.git
cd splitwise-mcp
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
Step 2: Add your API key
Create a file called .env in the project folder:
SPLITWISE_API_KEY=paste_your_key_here
Where to get your key:
- Go to secure.splitwise.com/apps
- Click "Register your application"
- Fill in any name (e.g., "My MCP Server")
- Copy the API Key into your
.envfile
Step 3: Connect to Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"splitwise": {
"command": "/full/path/to/venv/bin/python3",
"args": ["/full/path/to/splitwise_server.py"]
}
}
}
Restart Claude Desktop. You'll see a hammer icon — that means tools are connected.
Step 4 (Optional): Host on a Server for 24/7 Access
Want it running even when your laptop is off? Deploy to any VPS (I use Hostinger).
See DEPLOY.md for the full step-by-step guide.
Once hosted, connect Claude Desktop via:
{
"mcpServers": {
"splitwise": {
"command": "npx",
"args": ["-y", "mcp-remote", "http://YOUR_SERVER_IP:8000/mcp", "--allow-http"]
}
}
}
All Available Tools
| Tool | What It Does |
|---|---|
get_current_user |
Your Splitwise profile |
list_currencies |
All currencies Splitwise supports |
list_groups |
All your groups with member IDs |
get_group |
Group details — members, balances, debts |
list_expenses |
Recent expenses in a group |
get_expense |
Full details of one expense |
create_expense |
Add expense (equal, exact, or percentage split) |
update_expense |
Edit description, cost, currency, or date |
delete_expense |
Remove an expense |
create_payment |
Record a settlement between two people |
Troubleshooting
| Problem | Fix |
|---|---|
| Claude doesn't see Splitwise tools | Restart Claude Desktop after editing the config |
| 401 Unauthorized error | Check your API key in .env is correct |
| "No module named mcp" | Activate your venv: source venv/bin/activate |
| Server disconnected | Check server is running: sudo systemctl status splitwise-mcp |
| Wrong paths in config | Use full absolute paths, not relative ones |
Project Structure
splitwise-mcp/
├── splitwise_server.py ← The entire server (~500 lines, one file)
├── .env ← Your API key (never committed to git)
├── requirements.txt ← Python dependencies
├── architecture.md ← How everything connects
├── DEPLOY.md ← How to host on a VPS
└── README.md ← This file
Built With
- FastMCP — Python MCP framework
- Splitwise REST API v3
- httpx — HTTP client
- Hosted on Hostinger VPS (Ubuntu)
Built by a first-time Python developer who just wanted to stop manually logging trip expenses. If I can build it, you can use it.
License
MIT — use it however you like.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。