FinPlan Setup Guide¶
Connect your AI assistant to FinPlan's financial planning tools. Once connected, you can ask your assistant to calculate taxes, project investment growth, plan for goals, and more - and it will use real math instead of guessing.
FinPlan MCP Server URL: https://finplan-production-09c8.up.railway.app/mcp
Priming the agent to use FinPlan Sometimes agents need to be told to use tools. Making sure it knows about them often helps.
user> 'Do you have access to FinPlan tools for financial planning?'
agent> (inspecting) 'Yes, I do'.
ChatGPT¶
ChatGPT supports remote MCP servers through Developer Mode. Available for Pro, Plus, Team, Enterprise, and Edu plans.
Setup Steps¶
- Open chatgpt.com and sign in
- Go to Settings (gear icon)
- Navigate to Connectors > Advanced > Developer Mode and enable it
- Go back to Connectors and click Add new connector
- Fill in:
- Name:
FinPlan - MCP Server URL:
https://finplan-production-09c8.up.railway.app/mcp - Click Create
Using FinPlan in a Chat¶
- Start a new chat
- Click the Connectors button in the message composer
- Select FinPlan from the list
- Ask your question - ChatGPT will use FinPlan tools automatically
Things to Know¶
- Developer Mode is a beta feature. Memory is disabled while it's active.
- ChatGPT will ask you to confirm before running any tool that writes or modifies data.
- You need to enable the FinPlan connector in each new chat session.
References¶
Claude Web (claude.ai)¶
Claude on the web supports MCP connections through the integrations settings.
Setup Steps¶
- Go to claude.ai and sign in
- Click your profile icon in the bottom-left corner
- Select Settings
- Navigate to Integrations
- Click Add Integration
- Enter:
- Name:
FinPlan - URL:
https://finplan-production-09c8.up.railway.app/mcp - Save the configuration
- You have to click
Allow Allfor tool use to allow the tools.
Step 2 — Install the FinPlan skill
Skills give Claude domain knowledge about how to use FinPlan tools effectively (conventions, workflows, parameter formats).
- Download
skills.zipfrom the repo (or build it:cd skills && zip -r skills.zip .) - Go to Settings > Capabilities
- Scroll to Skills and ensure skills are toggled on
- Click Upload skill and select
skills.zip - Confirm the FinPlan skill appears in your skills list
Usage¶
Once connected, Claude will automatically have access to FinPlan tools. Just ask your question naturally:
- "Create a financial profile for someone born 1985-06-15, employed, making $95,000"
- "Project how $250,000 would grow over 30 years with moderate risk"
- "What's the long-term capital gains rate for $150,000 income filing married jointly?"
Claude Desktop¶
The Claude Desktop app supports MCP connections and skills. With the Cowork feature (requires Claude Max subscription), Claude can run multi-step financial planning tasks autonomously.
Setup Steps¶
Step 1 — Add the MCP connector
- Open Claude Desktop
- Go to Settings > Connectors
- Click Add connector and enter the remote MCP server URL:
- Save. You should see a tools icon confirming the connection.
Note: Remote MCP servers must be added via Settings > Connectors. They cannot be added through the
claude_desktop_config.jsonfile.
Step 2 — Install the FinPlan skill
- Download
skills.zipfrom the repo (or build it:cd skills && zip -r skills.zip .) - Open Settings > Capabilities
- Scroll to Skills and ensure skills are toggled on
- Click Upload skill and select
skills.zip - Confirm the FinPlan skill appears in your skills list
Usage¶
Start a new conversation or Cowork session. Claude will automatically use FinPlan tools when you ask about financial planning. For Cowork, describe a task like:
"Build me a retirement projection: I'm 35, earn $120k, contribute $500/mo to my 401(k) with a 50% employer match up to 6%. Show me projected outcomes at ages 55, 60, and 65."
The Desktop app must remain open while Cowork is running. Closing it ends the session.
Claude CLI¶
The Claude command-line tool supports MCP servers for hands-free financial planning from your terminal.
Configuration¶
Add the FinPlan server to your settings file.
Location: ~/.claude/settings.json (or project-level .claude/settings.json)
Install the FinPlan skill¶
Copy the skill files into your Claude Code skills directory:
Or if you have skills.zip, unzip it into the skills folder:
Claude Code auto-discovers skills from ~/.claude/skills/. You can also invoke it directly with /finplan.
Verification¶
Start Claude CLI and try a FinPlan command:
Claude Cowork (with Local File Storage)¶
Claude Cowork lets you persist your financial plans as local files, so you can save your plan in one session and pick it back up later.
Configuration¶
In your project's .claude/settings.json:
Workflow Example¶
-
Create your plan:
-
Save it:
-
Come back later and pick up where you left off:
Suggested File Organization¶
your-project/
├── .claude/
│ └── settings.json # MCP configuration
├── financial_plans/
│ ├── personal_2025.json # Your financial plan
│ ├── scenario_aggressive.json
│ └── scenario_conservative.json
└── ...
Gemini CLI¶
Google's Gemini CLI supports MCP servers. The Gemini web app (gemini.google.com) does not yet support custom MCP connections.
Install Gemini CLI¶
Configuration¶
Add the FinPlan server to your Gemini CLI settings.
Location: ~/.gemini/settings.json
Verification¶
Start Gemini CLI and try a FinPlan command:
Where Else Gemini Supports MCP¶
- Android Studio - Enable in File > Settings > Tools > Gemini > MCP Servers
- Firebase Studio - Configure in
.idx/mcp.json
References¶
Other AI Tools and Frameworks¶
OpenAI Agents SDK¶
If you're building agents with OpenAI's SDK, you can connect to FinPlan's MCP server directly:
from openai_agents import Agent, MCPServerSse
finplan = MCPServerSse(url="https://finplan-production-09c8.up.railway.app/mcp")
agent = Agent(
name="Financial Planner",
instructions="Help users with financial planning questions.",
mcp_servers=[finplan]
)
LangChain¶
from langchain_mcp import MCPToolkit
toolkit = MCPToolkit(server_url="https://finplan-production-09c8.up.railway.app/mcp")
tools = toolkit.get_tools()
AutoGen¶
from autogen import AssistantAgent
from autogen.tools import MCPToolWrapper
mcp_tools = MCPToolWrapper("https://finplan-production-09c8.up.railway.app/mcp")
assistant = AssistantAgent(
name="financial_planner",
llm_config={"model": "gpt-4"},
tools=mcp_tools.get_tools()
)
CrewAI¶
from crewai import Agent
from crewai_tools import MCPTool
finplan_tool = MCPTool(
name="finplan",
server_url="https://finplan-production-09c8.up.railway.app/mcp"
)
analyst = Agent(
role="Financial Analyst",
goal="Help users understand their financial projections",
tools=[finplan_tool]
)
Any HTTP Client¶
For any tool or script that can make HTTP requests:
# List all available tools
curl https://finplan-production-09c8.up.railway.app/mcp/tools
# Call a tool
curl -X POST https://finplan-production-09c8.up.railway.app/mcp/call \
-H "Content-Type: application/json" \
-d '{
"name": "calculate_federal_income_tax",
"arguments": {
"taxable_income_cents": 8500000,
"filing_status": "single",
"tax_year": 2025
}
}'
Troubleshooting¶
Can't Connect¶
- Make sure you're using the correct URL:
https://finplan-production-09c8.up.railway.app/mcp - Check your network connection and firewall settings
- Restart your AI assistant and try again
Tool Not Working¶
- Ask the agent to check if it has access to FinPlan tools.
- Double-check that you've enabled the FinPlan connector/integration in your settings
- Try asking a simple question first: "What filing statuses are available?"
- Make sure you're giving enough information - for example, tax calculations need an income amount, filing status, and tax year
Unexpected Results¶
All monetary values are in cents internally. If you see a number like 8500000, that's $85,000.00. Your AI assistant should translate this for you, but if something looks off, that's likely why.