Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Splitwise MCP Server Add $50 for dinner with John and split equally"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Model Context Protocol (MCP) server that integrates with Splitwise. Connect your AI assistant (Claude, Cursor, etc.) to manage Splitwise expenses using natural language — with voice support!
flowchart LR
Client[Claude / Cursor] -->|MCP| Server[splitwise-mcp]
Server -->|audio| Deepgram[Deepgram STT]
Deepgram -->|text| Gemini[Gemini 3 Flash]
Gemini -->|action| Splitwise[Splitwise API]Related MCP server: Splitwise MCP Server
Tool | Description |
| Send audio → Deepgram transcribes → Gemini processes → Splitwise executes |
| Send text → Gemini processes → Splitwise executes |
| Add expenses with support for groups, percentages, exclusions, and specific payers |
| Delete an expense by ID |
| List your Splitwise friends |
| Configure API credentials |
| Login with OAuth2 token |
Smart Name Matching: If Deepgram transcribes "Humeet" but your friend is "Sumeet", Gemini will ask for clarification instead of guessing.
Percentages: "Split 40% for me and 60% for Alice"
Groups: "Add to Apartment group" (Auto-fetches members)
Exclusions: "Add to Apartment but exclude Bob"
Payer: "Alice paid $50"
Deletion: "Delete expense 12345"
pip install splitwise-mcpClone the repository:
git clone https://github.com/hubshashwat/the-splitwise-mcp.git
cd the-splitwise-mcpCreate and activate a virtual environment:
python3 -m venv .venv
source .venv/bin/activateInstall the package:
pip install -e .You'll need API keys depending on which features you want:
Splitwise API Keys (https://secure.splitwise.com/apps/new)
Register a new application
Get: Consumer Key, Consumer Secret, and API Key
Required for: add_expense, list_friends, delete_expense
Gemini API Key (https://aistudio.google.com/) - Optional
Create API key (free tier available)
Model: Uses Gemini 3.0 Flash - ensure your API key has access to this model
Required for: text_command (natural language processing)
Deepgram API Key (https://console.deepgram.com/) - Optional
Sign up and get API key (free tier available)
Required for: voice_command (audio transcription)
Summary:
Text-only users: Need Splitwise + Gemini keys (skip Deepgram)
Voice users: Need all 5 keys
Basic API users: Only need 3 Splitwise keys
Set environment variables in your shell or add to your Claude Desktop config (see below).
You can use this server in two ways:
Run the voice/text agent directly in your terminal:
# Install the package
pip install splitwise-mcp
# Download the agent script
curl -O https://raw.githubusercontent.com/hubshashwat/the-splitwise-mcp/main/run_agent.py
# Set environment variables
export SPLITWISE_CONSUMER_KEY="your_key"
export SPLITWISE_CONSUMER_SECRET="your_secret"
export SPLITWISE_API_KEY="your_api_key"
export GEMINI_API_KEY="your_gemini_key"
export DEEPGRAM_API_KEY="your_deepgram_key"
# Run the agent
python run_agent.pyCommands:
v or voice - Record 10 seconds of audio and process it
t or text - Type your command
q or quit - Exit
Example session:
🤖 Splitwise Voice Agent
Enter command (voice/text/quit): t
Enter request: Add expense of $50 with John for dinner
⚠️ Proposed Action:
Function: add_expense
Args: {
"description": "dinner",
"cost": 50.0,
"split_with": ["John"]
}
Proceed? (yes/edit/cancel): yes
✅ Expense added!This server uses stdio transport and works with all MCP-compatible clients:
Add to your config:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"splitwise": {
"command": "full path of python",
"args": ["-m", "splitwise_mcp.server"],
"env": {
"SPLITWISE_CONSUMER_KEY": "your_consumer_key",
"SPLITWISE_CONSUMER_SECRET": "your_consumer_secret",
"SPLITWISE_API_KEY": "your_api_key",
"GEMINI_API_KEY": "your_gemini_key",
"DEEPGRAM_API_KEY": "your_deepgram_key"
}
}
}
}Note: If
splitwise-mcpconsole command is available, you can use"command": "splitwise-mcp"without args instead.
Then in Claude: "Add an expense of $50 with John for dinner"
Use the same config format with claude-cli --mcp-config.
Add to your MCP settings (.vscode/settings.json or Cursor settings):
{
"mcp.servers": {
"splitwise": {
"command": "full path of python",
"args": ["-m", "splitwise_mcp.server"],
"env": {
"SPLITWISE_CONSUMER_KEY": "your_consumer_key",
"SPLITWISE_CONSUMER_SECRET": "your_consumer_secret",
"SPLITWISE_API_KEY": "your_api_key",
"GEMINI_API_KEY": "your_gemini_key",
"DEEPGRAM_API_KEY": "your_deepgram_key"
}
}
}
}Then you can ask your AI assistant: "Use Splitwise to add an expense..."
This server is compatible with any MCP client supporting stdio transport. Use the same configuration pattern.
Note: If you installed from source instead of pip, use the full path to the executable:
macOS/Linux: "/path/to/the-splitwise-mcp/.venv/bin/splitwise-mcp"
Windows: "C:\\path\\to\\the-splitwise-mcp\\.venv\\Scripts\\splitwise-mcp.exe"
To run the MCP server over HTTP for remote clients:
.venv/bin/uvicorn splitwise_mcp.sse:app --host 0.0.0.0 --port 8000Connect via: http://YOUR_IP:8000/sse
Run tests:
.venv/bin/python tests/test_logic.pyIf the agent says "Recording finished" immediately but captures no audio (Volume: 0.0), your terminal likely lacks microphone permission.
Go to System Settings > Privacy & Security > Microphone.
Enable access for your terminal app (Terminal, iTerm, VS Code, etc.).
Restart your terminal for changes to take effect.
MIT
Add an expense to Splitwise, supporting unequal splits, groups, and precise control.
Args:
amount: The total cost (e.g., "70", "10.50").
description: A brief description.
friend_names: Friends to split with. Can be empty if using group_name.
split_map: Optional dict for unequal splits. Keys=Names (or 'me'), Values=Amount/Percentage.
Example: {'me': '40%', 'Alice': '60%'} or {'me': '10', 'Bob': '20'}
group_name: Optional group to add expense to.
payer_name: Optional name of who paid. Defaults to 'me'.
exclude_names: Optional list of names to exclude from a group split.
| Name | Required | Description | Default |
|---|---|---|---|
| amount | Yes | ||
| description | Yes | ||
| friend_names | Yes | ||
| split_map | No | ||
| group_name | No | ||
| payer_name | No | ||
| exclude_names | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description bears full burden. It discloses core functionality and split options but omits details on error handling, permissions, or side effects. The examples for split_map are helpful but do not cover all potential issues.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is well-organized with an Args section and examples. It is concise yet informative, though the verbose Example formatting could be slightly more compact.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the number of parameters (7) and no annotations, the description covers most parameters and their relationships. The presence of an output schema likely covers return values. Minor gaps include lack of error scenarios and ordering constraints.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With 0% schema coverage, the description compensates well: it explains amount as 'total cost', description as 'brief description', and provides explicit examples for split_map. However, some parameters like exclude_names lack contextual detail about how they interact with group splits.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's action: 'Add an expense to Splitwise', and highlights key features like unequal splits and groups. It distinguishes itself from siblings like delete_expense and list_friends.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description includes parameter explanations and usage notes (e.g., friend_names can be empty if using group_name, split_map examples). It lacks explicit comparison to alternative tools but provides sufficient context for correct invocation.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Configure the Splitwise client with API credentials. You must provide either (consumer_key AND consumer_secret) OR api_key.
| Name | Required | Description | Default |
|---|---|---|---|
| consumer_key | No | ||
| consumer_secret | No | ||
| api_key | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must carry the burden of behavioral disclosure. While 'configure' implies a mutation, the description fails to specify whether credentials are validated, persisted, or if reconfiguration overwrites previous settings. This lack of detail leaves the agent uncertain about side effects.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is extremely concise, consisting of two short sentences. The critical constraint is front-loaded in the first sentence, and the second sentence adds essential usage guidance. Every word earns its place with no redundancy.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description is adequate for a simple configuration tool but lacks information about whether this step is a prerequisite for other tools, or what the output schema contains. Since an output schema exists, the description could note that a confirmation or status is returned. More context would make it complete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With schema description coverage at 0%, the description adds value by explaining the OR relationship between parameters. However, it does not explain the format or purpose of each parameter beyond their titles. For a configuration tool, more detail (e.g., where to obtain these) would improve semantics.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Configure the Splitwise client with API credentials.' The verb 'configure' and noun 'Splitwise client' are specific, and the tool is distinctly different from siblings like 'add_expense' or 'list_friends', which handle data operations.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides explicit instructions on required inputs: 'You must provide either (consumer_key AND consumer_secret) OR api_key.' This clarifies the parameter dependency. However, it does not mention when to use the tool versus alternatives, though siblings are unrelated to configuration, so this is minor.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Delete an expense by its ID.
| Name | Required | Description | Default |
|---|---|---|---|
| expense_id | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided, and the description does not disclose behavioral traits like whether deletion is permanent, requires authentication, or has side effects. The burden falls entirely on the description.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
One sentence, no fluff. The description is front-loaded and concise, fitting for a simple tool.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity (one required param, output schema present), the description is minimally complete. However, it lacks behavioral details and usage context, leaving gaps for an AI agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The single parameter 'expense_id' has no schema description (0% coverage), and the description adds no extra meaning beyond 'by its ID'. No format, example, or constraints are provided.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action (delete) and resource (expense), specifying deletion by ID. It effectively distinguishes from siblings like 'add_expense'.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool versus alternatives, such as other expense management tools. Lacks context like prerequisites or scenarios where deletion is appropriate.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List all friends of the current user on Splitwise. Returns a formatted string list of friends.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Discloses return type as 'formatted string list' but no further details on formatting or behavior. No annotations to supplement, so adequate but minimal.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two concise sentences, no wasted words, front-loaded with purpose.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Sufficient for a simple parameterless tool with output schema noted. Could mention empty list handling but not critical.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
No parameters; schema coverage 100%. Description adds no param info but not required.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Clearly states the verb 'List', resource 'friends', and scope 'of the current user'. Distinguishes from siblings which are expense and configuration tools.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool vs alternatives. Missing context for tool selection.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Log in using an existing OAuth2 Access Token. Useful for integrations where authentication is handled externally (e.g. ChatGPT).
| Name | Required | Description | Default |
|---|---|---|---|
| access_token | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided, so the description carries full burden. It only states the action but does not disclose important behaviors such as whether it replaces an existing session, validates the token, or any side effects. The return value is also not mentioned despite an existing output schema.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, no extraneous words, and front-loads the action in the first sentence. Every part is useful.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The tool is simple with one parameter and no nested objects. An output schema exists, so return details are not needed. However, the description lacks prerequisites (e.g., token must be valid, from an external OAuth provider) and what happens post-login. It is adequate but not fully complete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 0% description coverage, and the description only mentions 'existing OAuth2 Access Token' without explaining parameter details like format, source, or constraints. It adds minimal value beyond the schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool logs in using an existing OAuth2 Access Token, and it distinguishes itself from sibling tools (no other authentication tools) by specifying a use case for external integrations.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly advises when to use the tool ('integrations where authentication is handled externally, e.g., ChatGPT'), providing clear context. However, it does not mention when not to use or suggest alternatives, but siblings do not include other auth tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Process a text command for Splitwise.
Interprets natural language text using Gemini and executes Splitwise actions. Use this when you already have text (e.g., from a chat message) instead of audio.
Args: text: Natural language command (e.g., "Split 50 with Sumeet for dinner").
Returns: The result of the command (e.g., confirmation, clarification request, or error).
| Name | Required | Description | Default |
|---|---|---|---|
| text | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It states that the tool 'executes Splitwise actions' and returns results, implying mutation, but it does not specify authentication requirements, side effects, or the scope of actions possible. This is adequate but not fully transparent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is extremely concise with only three sentences plus Args/Returns, every sentence adds value. The key purpose is front-loaded in the first sentence, and the structure is clean and efficient.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity as a natural language command processor and the presence of an output schema, the description covers input and output adequately. It provides a clear example and distinguishes from audio input. However, it could mention the AI model (Gemini) and the range of possible actions for completeness.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 0% description coverage, so the description must compensate. It adds meaning by describing the 'text' parameter as a 'Natural language command' with a concrete example ('Split 50 with Sumeet for dinner'), which helps the agent understand usage. However, it could provide more format guidance.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool processes text commands for Splitwise by interpreting natural language and executing actions. It explicitly distinguishes itself from the sibling 'voice_command' by noting this version is for text input instead of audio, making the purpose unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives explicit context for when to use this tool ('when you already have text instead of audio'), which differentiates it from the sibling 'voice_command'. However, it does not mention when not to use it or provide guidance on alternatives among other siblings for specific tasks, missing some depth.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Process a voice command for Splitwise.
Accepts base64-encoded audio (WAV or MP3 format), transcribes it using Deepgram, processes the intent using Gemini, and executes Splitwise actions.
Args: audio_base64: Base64-encoded audio data (WAV or MP3).
Returns: The result of the voice command (e.g., confirmation, clarification request, or error).
| Name | Required | Description | Default |
|---|---|---|---|
| audio_base64 | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries full burden. It discloses the high-level process (transcription, intent, execution) and possible returns (confirmation, clarification, error). However, it does not detail potential side effects like data modification or required permissions.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is front-loaded with purpose and efficiently explains the process. Minor redundancy (first sentence then restates) but overall well-structured and concise.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Despite complexity (voice pipeline with external services), the description covers the process, input, and return types. Mentions Deepgram and Gemini. Lacks details on audio size limits or latency, but still fairly complete given output schema exists.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema has only one parameter (audio_base64) with no description. The description adds meaning by specifying format constraints (WAV or MP3) and encoding. Given 0% schema coverage, the description compensates well.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states that the tool processes voice commands for Splitwise by accepting base64 audio, transcribing it, processing intent, and executing actions. It distinguishes itself from siblings like text_command by explicitly focusing on voice input.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool versus alternatives such as text_command. The description does not mention prerequisites, when voice is appropriate, or when not to use it.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.2add_expenseconfigure_splitwisedelete_expenselist_friendslogin_with_tokentext_commandvoice_commandScored across 7 tools
Each tool has a clearly distinct purpose: authentication setup, login, adding/deleting expenses, listing friends, and text/voice command processing. No overlapping functionality.
Most tools follow a verb_noun pattern (add_expense, delete_expense, list_friends), but login_with_token and configure_splitwise are slightly different, and text_command/voice_command are adjective_noun. Minor inconsistency but still readable.
Seven tools cover essential Splitwise actions like authentication (configure, login), expense management (add, delete), friend listing, and natural language interfaces, which is well-scoped for the server's purpose.
Notable missing operations: no update or get/list expenses (only add/delete), no group management beyond expense addition, and no tool to retrieve expense details. The command tools may partially compensate but lack explicit coverage.
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yo…
A Model Context Protocol server for Wix AI tools
Hosted MCP server for LLM cost estimation, model comparison, and budget-aware routing.
The Mercado Pago MCP Server implements the Model Context Protocol to provide AI agents and LLMs with access to Mercado Pago's APIs and tools within compatible development environments. It acts as an intermediary that translates Mercado Pago resources into executable functions (tools) that AI applications can invoke to perform actions and automate flows. The server simplifies integration, enables using documentation to implement or improve code, and optimizes operations through natural language interactions without manual implementations.
