Guide

How to connect TallyPrime to Claude or ChatGPT (MCP)

MCP lets an AI assistant read from tools you connect, with your permission. Connect your TallyPrime books once and you can ask Claude or ChatGPT who owes you the most, in plain words.

Updated 22 October 2026 · By the TallyNexa team

Short answer: Install the TallyNexa connector so your TallyPrime books sync to TallyNexa, then add https://tallynexa.ai/mcp as a custom connector in Claude (Settings > Connectors) or ChatGPT (developer mode), sign in to TallyNexa and choose which books the assistant may read. The assistant reads as you, within your role, and any change it proposes waits in TallyNexa until you say yes there.

What MCP is

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to outside tools and data. An MCP server offers a set of actions, such as listing overdue customers; the assistant calls them when your question needs them.

For Tally, that means the assistant asks for figures from your books instead of guessing, and you decide which books it may see.

Before you start

  1. Install the TallyNexa connector on the Windows computer that runs TallyPrime (tallynexa.ai/download) and link it to your account.
  2. Open the companies in TallyPrime and choose the books TallyNexa may read.
  3. Let the first sync finish. It goes month by month; a book of about 90,000 vouchers takes minutes.

The MCP address is https://tallynexa.ai/mcp. It is the same for everyone; signing in decides whose books it reads.

Claude (claude.ai or Claude Desktop)

  1. Open Settings > Connectors.
  2. Choose Add custom connector and paste https://tallynexa.ai/mcp.
  3. Choose Connect. A TallyNexa sign-in page opens.
  4. Sign in, choose the books Claude may read, and whether it may propose changes.
  5. Choose Allow.

Start a new chat and ask: "Who owes me the most, and since when?" If a menu looks different, check Claude’s help; menus and plan requirements change over time.

ChatGPT

  1. Open Settings > Apps and connectors > Advanced and turn Developer mode on.
  2. Choose Create, give it a name, paste https://tallynexa.ai/mcp as the address and pick OAuth for sign-in.
  3. Sign in to TallyNexa when asked, choose the books, and choose Allow.

OpenAI changes these menus, and which plans can add custom connectors. If a step looks different, check OpenAI’s help for the current route.

Claude Code

From a terminal, add the server, then sign in when prompted:

claude mcp add --transport http tallynexa https://tallynexa.ai/mcp

If you prefer a key to signing in, make one in TallyNexa under Preferences > Assistants > Make a key and use it with the same address.

What the assistant can and cannot do

  • It reads as you. It sees only the books you chose and only what your role in TallyNexa allows.
  • It can propose a change. The proposal waits in TallyNexa as a card until you say yes there; nothing is posted from the chat.
  • On the Insight plan, nothing an assistant proposes is posted to Tally.
  • You can end the connection at any time.

Questions to try

  • "Which customers are more than 90 days overdue, and how much do they owe?"
  • "Compare this month’s sales with the same month last year."
  • "What were our five biggest expenses this quarter?"
  • "Which suppliers do we owe the most, and when is it due?"

Check important figures before you act on them, as you would with any report. If you would rather stay inside TallyNexa, Nexa AI answers questions from the same books and marks each figure as read from the book or worked out.

The open-source route

There are also open-source MCP servers for Tally on GitHub. They run on your own computer with Node.js, talk to TallyPrime directly on port 9000, and let a desktop assistant query Tally live. No other service sits between Tally and your assistant.

In return, you install and update the server yourself, Tally must be open and reachable from wherever the server runs, and the assistant can reach your books only while that computer is on. Pick it if you are comfortable looking after that set-up. Pick a hosted connection if you want it to work from any device and any assistant without running a server.

Which books to share

Share only what the question needs. If you run several companies, start with one, see how the answers look, and add the rest later. You can change the books an assistant may read from TallyNexa without touching the assistant.

Questions people ask

Can ChatGPT read my Tally data?

Yes, through an MCP connector. Add https://tallynexa.ai/mcp in ChatGPT’s developer mode, sign in to TallyNexa and choose the books it may read. Your books reach TallyNexa through its connector on the computer that runs TallyPrime.

Tally ko ChatGPT ya Claude se kaise connect kare?

Install the TallyNexa connector and let the books sync, then add https://tallynexa.ai/mcp as a custom connector in Claude or ChatGPT, sign in to TallyNexa and allow access.

Can the assistant change my Tally books?

Not on its own. Any change it proposes waits in TallyNexa as a card until you say yes there, and on the Insight plan nothing it proposes is posted to Tally.

Is an open-source Tally MCP server an option?

Yes, if you are comfortable running a Node.js server on a computer that can reach TallyPrime on port 9000. It works while that computer and Tally are on.

What is the TallyNexa MCP address?

https://tallynexa.ai/mcp. You sign in to TallyNexa when you add it.

In TallyNexa

More guides

Built for Tally.
Ready today.