The Tally MCP server is currently in beta and subject to change.
What is MCP?
The Model Context Protocol (MCP) lets you build Tally forms and retrieve forms or submissions
using natural language through AI assistants like Claude.
Server URL
The Tally MCP server is available at:
Authentication
The Tally MCP server supports two authentication methods:
Option 1: OAuth (Recommended)
The easiest way to connect is through OAuth. Many AI assistants support OAuth authentication for MCP servers, allowing you to connect with just a few clicks without manually managing API keys.
When your AI assistant prompts for OAuth authentication:
- Click the connect button or authorization link
- Log in to your Tally account (if not already logged in)
- Authorize the connection
- The AI assistant will automatically configure the connection
Option 2: API Key
You can also authenticate using an API key in the Authorization header:
Authorization: Bearer tly-xxxx
Learn more about how to get your API key.
Configuration
Claude Desktop
This method works for Claude Desktop and claude.ai, and applies across all Claude apps that support connectors, including Claude Chat, Claude Code, and Claude Cowork.
- Open Claude Desktop (or claude.ai in your browser) and navigate to Customize > Connectors > Add custom connector from the sidebar on the left.
- Give the connector a name and the Tally MCP server URL.
- Click “Add” to save the connector.
- Restart Claude Desktop
- The first time you use the Tally MCP server, Claude will open an OAuth flow to authenticate with your Tally account.
Claude Code CLI
To connect Claude Code to the Tally MCP server, run the following command in your terminal:
claude mcp add tally --transport http https://api.tally.so/mcp
Claude Code will handle the OAuth flow automatically when you first use the Tally MCP server. You can verify the connection by running the /mcp command inside Claude Code.
Cursor
- Open Cursor Settings > Tools & MCP
- Click New MCP Server
- Paste the following JSON to configure the Tally MCP server:
{
"mcpServers": {
"tally": {
"url": "https://api.tally.so/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY_HERE"
}
}
}
}
Example Usage
Once configured, you can ask your AI assistant to create forms using natural language. Here are some examples:
Create a simple contact form
The AI assistant will automatically create a contact form with commonly expected fields based on its understanding of what a typical contact form should include.
If the generated form isn’t exactly what you wanted, you can easily modify it by following up:
We don't need the Company field, remove the consent checkbox and maybe add a captcha
The AI assisant should correctly modify the form and save it.
Alternatively, you can be explicit about exactly what fields you want:
Create a contact form with these specific fields:
- Name (required)
- Email (required)
- Phone (optional)
- Message (required)
- Captcha (required)
The AI assistant should create a form with the specified fields and save it.
Update this form https://tally.so/r/a1B2c3
- Remove the phone number field
- Also add a captcha
- Make company field optional
The AI assistant will use the MCP tools to load your form, make the requested changes, and save the updated version.
You can fetch submissions for any form, with optional filters for status and date range:
Show me all completed submissions for https://tally.so/r/a1B2c3 from January 2025
The AI assistant will fetch and display the matching submissions for your form, including question labels and response data.
You can list forms across your accessible workspaces and ask for filtering by name, status, or recency:
Show my 10 most recently updated published forms
The AI assistant will call the MCP list forms tool and return matching forms with key details such as form name, URL, status, and last updated date.
Example usage with Claude
Configuring the Tally MCP server using OAuth with Claude