- UPDATED
- ENFORCED
- REVIEW
- AUDITED
Short answer
How can I update my expense policies using AI?
You describe the changes you want to make to your categories, tags, or rules in plain English. Your AI then executes those updates directly in Expensify. You'll see the specific policy changes and new rules applied immediately.
Policy outcomes
Where your changes land.
Your AI handles the technical configuration of your expense setup.
UPDATED
New categories and tags
The AI modifies your existing expense categories or adds new tags based on your instructions. You get a clean, updated list of labels for your team to use.
ENFORCED
Automated expense rules
Tell your agent to set new spending limits or restrictions. It creates the logic needed to flag or block specific expenses automatically.
REVIEW
Human oversight required
The AI proposes the logic for complex policy changes. You must review and confirm any major shifts in spending rules before they go live.
AUDITED
Policy list visibility
Your agent pulls your current policy configurations so you can see exactly what is active. This ensures your instructions match your existing setup.
The workflow
What your AI does when the request arrives.
The agent translates your natural language instructions into specific Expensify configuration changes.
Fetch current settings
The agent looks up your existing expense policies to understand the current structure and constraints.
get_policy_listModify policy details
The agent updates specific categories or tags within your existing policy framework.
update_policyApply new logic
The agent writes new rules to govern how expenses are flagged or approved.
create_expense_rulesVerify changes
The agent checks the updated policy list to ensure the new rules and categories are correctly applied.
get_policy_list
Try it
Copy these to start.
Use these prompts to trigger policy updates in Expensify.
Starting points
Swap the bracketed text for your actual category names or specific spending limits.
Expensify Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_ie4EE7sHW9cYpZclk8tSw8uSPp4wcF4Dnc6qnBRo/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Expensify capabilities are ready to use.
{
"mcpServers": {
"expensify-mcp": {
"url": "https://edge.vinkius.com/vk_preview_ie4EE7sHW9cYpZclk8tSw8uSPp4wcF4Dnc6qnBRo/mcp"
}
}
}Add a new expense category called [Travel] to my current policy.
Create a rule that flags any expense over $[500] for manual review.
Update the [Marketing] tag to include [Social Media Ads].
Show me all the current expense rules in my policy list.
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Expensify once, then ask.
Log in once to link your account. Your credentials stay encrypted, and you can immediately start managing policies through your preferred AI client.
Connect Expensify to your AIFAQ
How this task behaves.
- 01
Can the AI delete my entire expense policy?
No. The AI only has tools to update existing policies or create new rules. It cannot delete a policy entirely.
- 02
Will the AI automatically approve expenses?
No. The AI can create rules to flag expenses, but it does not have a tool to perform the actual approval of individual expense reports.
- 03
Can I see what the AI is about to change?
Yes. You should always ask your agent to show you the current policy list before it makes updates so you can verify the context.
- 04
Does the AI handle employee reimbursements?
No. This MCP is specifically for managing the rules, categories, and tags that govern how expenses are handled.
- 05
Can the AI add new users to Expensify?
No. The available tools are limited to policy management, categories, and expense rules.
More questions about Expensify? The Connector page answers them. See everything the Expensify Connector can do
Connect Expensify to Claude, Cursor, ChatGPT & more
