india-stack-mcp
Enables AI agents to manage Indian enterprise systems by providing tools for accounting (TallyPrime), invoicing (Zoho Books), and GST compliance (validation, e-invoice generation, and GSTR-2B reconciliation), with dry-run-first writes and an audit trail.
README
india-stack-mcp
Production-grade MCP servers for Indian enterprise systems. Agents that can do books and file taxes in India — starting with TallyPrime, the accounting system behind millions of Indian SMEs.
Why
Frontier agents can reason about GST, but they cannot touch Tally. There is no modern API, no OAuth, no SDK — just a local XML gateway that punishes the unprepared. This project is the missing connector layer: typed tools, dry-run-first writes, idempotency, human approval hooks, and a tamper-evident audit trail, packaged as standard MCP servers that work in Claude Desktop, Claude Code, Cursor, Codex, or any MCP client.
Status
| Connector | Scope | State |
|---|---|---|
tally-mcp |
Companies, ledgers, vouchers (7 types), day book, trial balance | alpha |
zoho-mcp |
Organizations, contacts, invoices, items, payments, chart of accounts | alpha |
gst-mcp |
GSTIN validation, e-invoice (IRN) payloads + NIC-style encryption, GSTR-2B reconciliation | alpha |
Quick start
npm install
npm run build
Point Claude Desktop / Cursor at the built server (see
examples/claude_desktop_config.json):
{
"mcpServers": {
"tally": {
"command": "node",
"args": ["<abs-path>/packages/connector-tally/bin/tally-mcp.js"],
"env": {
"TALLY_URL": "http://localhost:9000",
"TALLY_COMPANY": "Your Company Pvt Ltd"
}
},
"zoho": {
"command": "node",
"args": ["<abs-path>/packages/connector-zoho/bin/zoho-mcp.js"],
"env": {
"ZOHO_CLIENT_ID": "...",
"ZOHO_CLIENT_SECRET": "...",
"ZOHO_REFRESH_TOKEN": "...",
"ZOHO_DC": "in",
"ZOHO_ORGANIZATION_ID": "..."
}
}
}
}
Requirements on the Tally side: TallyPrime running locally with XML serving enabled on
port 9000 (F1 → Settings → Connectivity). See docs/tally-xml-notes.md.
Zoho Books needs a self-client OAuth refresh token with Books scopes
(ZohoBooks.fullaccess.all or narrower), created at the Zoho API Console. Credentials can
also live in a file-based vault instead of env vars (CredentialVault in core).
Tools (27)
Tally (12) — Reads: tally_list_companies, tally_get_company, tally_search_ledgers,
tally_get_ledger, tally_day_book, tally_trial_balance, tally_ledger_balances,
tally_list_voucher_types, tally_find_vouchers. Writes: tally_create_ledger,
tally_create_voucher (7 voucher types), tally_delete_voucher.
Zoho Books (11) — Reads: zoho_list_organizations, zoho_list_contacts,
zoho_get_contact, zoho_list_invoices, zoho_get_invoice, zoho_list_items,
zoho_chart_of_accounts. Writes (dry-run by default): zoho_create_contact,
zoho_create_invoice, zoho_create_item, zoho_create_customer_payment.
GST ops (4) — gst_validate_gstin (mod-36 checksum, state, PAN/entity type),
gst_einvoice_build (canonical v1.01 payload with tax splits + round-off),
gst_einvoice_generate (IRN via pluggable IRP client; NIC-style encrypted envelopes:
RSA-wrapped SEK + AES-256-CBC + HMAC-SHA256), gst_reconcile_2b
(matched/mismatch/missing classification across your purchase register vs GSTR-2B,
with ITC-at-risk quantification).
Every write returns the exact upstream request body as a preview until you pass
dry_run: false. Retries with the same idempotency_key never double-post.
Safety model
- Writes default to preview; committing requires explicit opt-in.
- Vouchers are balance-checked client-side before any network call.
- Every invocation appends a redacted, hash-chained audit record (
INDIA_STACK_AUDIT_LOG). - Destructive tools can require human approval via a pluggable
Approver.
See docs/architecture.md for the full design.
HTTP mode
npm run start -w @india-stack/demo-host # POST http://localhost:8787/mcp/tally
Development
npm run test # unit + golden-file tests (no Tally needed)
TALLY_LIVE=1 npm run test:live # hits a real local TallyPrime
npm run lint && npm run typecheck && npm test
License
Apache-2.0
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.