Tijori Finance

What you can do

"Pull Zomato's last 6 quarters of revenue and EBITDA margin, then read theirlatest concall transcript — are they hitting the targets management guided for?"
"Bajaj Finance's FII stake dropped 400bps last year while promoters held flat —pull their latest investor presentation to figure out if this is distribution or passive rotation."
"Crude is up — check the raw materials data for chemical spreads, then pullrevenue mix and operational metrics for Deepak Nitrite to see if margins are at risk."

"The goal is to turn data into information, and information into insight." — Carly Fiorina

In Action

1 — Cross-tool deep dive: Titan

Revenue mix pulled first, then cross-referenced against P&L data to explain trends

[Titan — Revenue Mix × P&L]

2 — IndiGo: P&L history

Full income statement history retrieved in one call — revenue, EBITDA, PAT across years

[IndiGo — P&L History]

3 — Concall → Financial projection

Knowledge base fetched the latest earnings call transcript, Claude read it and built a forward projection model

[Concall → Financial Projection]

Why this exists

Most financial MCP servers are built for US markets — Yahoo Finance, SEC filings, S&P data.

Nothing existed for India.

Tijori Finance is the most comprehensive source for Indian equity data — operational KPIs, revenue segment breakdowns, market share trends, and curated investor documents that aren't available anywhere else. This MCP server exposes all of it to Claude.

Tools

v1 — Stable

v2 — Stable

What's New

2026-06-12 — Screener overhaul

• Alternate data queries. The Advanced Screener's second box now works ad-hoc: screen_companies({ alternate: "market share > 50" }) or { alternate: "revenue from Defence > 50", filters: "Market Capitalization > 1000" }. Relationships: makes, revenue from, market share, uses, caters to, has plant in, combinable with AND/OR/NOT.

• The three screener checkboxes. latest_results_only (only companies that reported the latest quarter), superstar_investors (only companies a whale investor holds, with the holder named per row), and sme (search the SME-listed universe).

• Popular screens actually run now. list_popular_screens previously returned queries with the AND glued to the next field name (newlines in the page's links were being stripped), and dropped each screen's category, description, market-share query, and superstar-investor flag. It now returns all of those, and screen_companies { preset: "Monopoly Companies" } runs any of them through the same endpoint the website uses — including the ones advanced queries can't express.

• The advanced screener exposes Tijori's full query language. Queries support every field in Tijori's catalog, % values (ROCE > 20%), comparisons between fields (Net Sales > 3Yrs ago Net Sales), and arithmetic (capex/Net Block > 0.5).

• New search_screener_fields tool. Tijori has ~1,500 financial metrics once you count time-period variants (3yr Avg ROCE, 5yr Growth Net Sales, 10Yrs ago PAT…). Search the catalog to find the exact name, then use it in a query.

• Result caps + pagination. Screens used to return every matching row (1,000+ for loose queries). Results are now capped at 50 by default (limit raises it) and offset pages through the rest — "show me the next 50" is served instantly from cache without re-querying Tijori. total_results always reports the full count.

2026-06-10 — Reliability & performance

• Long, multi-tool queries no longer time out. Asking for a full workup in one go — e.g. "build a model for MTAR: overview + financials + concall + revenue mix + operational KPIs" — used to overwhelm the browser and fail with page.goto: Timeout errors. Page loads are now faster (ads, trackers and images are skipped), capped to a safe number of concurrent loads, and retried once on a transient timeout. Tools also wait for the actual data to render rather than a fixed delay, so you get complete results instead of half-loaded ones.
• Fixed a startup browser leak. A burst of parallel tool calls on a cold start could spin up several Chromium instances and leave most of them orphaned. It now launches exactly one and shares it.
• get_market_share returns cleanly when there's no data. For companies that simply have no market-share metrics (most non-lenders), it now returns an empty result with a note instead of erroring — which previously made the assistant retry the same call in a loop.

Setup

Requirements

• Node.js v18+ — install the LTS version
• A Tijori Finance account — free account works but has data limits; Pro recommended for full access
• Claude Desktop

Option 1 — Wizard (recommended)

The setup script handles everything automatically: installs packages, downloads the browser, authenticates, and writes the Claude Desktop config for you.

Windows

1. Download Node.js and install it (choose LTS)
2. Download Claude Desktop and install it
3. Download this repo and unzip it anywhere
4. Double-click setup.bat inside the folder
5. Follow the prompts — enter your Tijori email/password, then log in through the browser window that opens
6. Fully quit and reopen Claude Desktop

Mac

1. Download Node.js and install it (choose LTS)
2. Download Claude Desktop and install it
3. Download this repo and unzip it anywhere
4. Double-click setup.command inside the folder — if macOS blocks it ("unidentified developer"), right-click the file and choose Open the first time
5. Follow the prompts — enter your Tijori email/password, then log in through the browser window that opens
6. Fully quit and reopen Claude Desktop (Cmd+Q, not just closing the window)

Linux (or any terminal)

git clone https://github.com/LaZZy0v0/tijori-finance-mcp.gitcd tijori-finance-mcpnode setup.js

Follow the prompts, then fully quit and reopen Claude Desktop.

Option 2 — Manual

For those who want to see exactly what each step does.

1. Clone and set credentials

# Mac / Linuxgit clone https://github.com/LaZZy0v0/tijori-finance-mcp.gitcd tijori-finance-mcpcp .env.example .env
# Windowsgit clone https://github.com/LaZZy0v0/tijori-finance-mcp.gitcd tijori-finance-mcpcopy .env.example .env

Open .env in any text editor and fill in your Tijori credentials:

[email protected]TIJORI_PASSWORD=yourpassword

2. Install packages and browser

npm installnpx playwright install chromium

npm install downloads the Node.js dependencies. playwright install chromium downloads a ~150 MB Chromium browser used to maintain your Tijori session — one-time only.

3. Authenticate

node discover.js

A browser window opens at the Tijori Finance sign-in page. Log in as you normally would. Once you're in, the script visits a few pages in the background to capture API endpoints, then the window closes automatically. Your session is saved to output/session.json.

4. Configure Claude Desktop

Find your config file and add the block below. The path to src/index.js must be absolute.

{  "mcpServers": {    "tijori-finance": {      "command": "node",      "args": ["C:/Users/yourname/tijori-finance-mcp/src/index.js"]    }  }}

5. Fully quit and reopen Claude Desktop

Test it

"Search Tata Steel using Tijori MCP"

If Claude returns a result, you're connected.

Session expiry

Tijori sessions expire periodically. When tools stop working:

npm run reauth

A browser window opens — log in manually and you're back.

How it works

The server uses Playwright to maintain an authenticated browser session with Tijori Finance. Each tool navigates to the relevant page or calls the underlying API, parses the response, and returns structured JSON to Claude. Results are cached in-memory (6 hours for financials, 30 minutes for metrics) to keep things fast.

[Architecture]

Disclaimer

This project is not affiliated with Tijori Finance. It requires your own paid Tijori Finance subscription. Use for personal research only — do not redistribute the underlying data.