Autonomous trading MCP for T-Bank Invest
Tinkoff Trader MCP
An MCP server for T-Bank Invest with market analysis, backtesting five strategies, risk management and order placement, with live trading off by default
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Once TINKOFF_LIVE_TRADING is explicitly enabled, the server places and cancels real exchange orders with account money
- Start with TINKOFF_MODE=sandbox and a read-only token, and enable live trading deliberately, one trade at a time
Install
Manual install
pip install tinkoff-trader-mcpInstall from PyPI; TINKOFF_TOKEN and TINKOFF_ACCOUNT_ID are needed afterward.
This is third-party code. Review the repository files before installing.
What it does
A Python server provides 12 tools for the Russian stock market: a market scan with regime classification (trending, ranging, volatile), momentum, RSI, ADX, deep single-instrument analysis, Engle-Granger cointegrated pair search, backtesting five built-in strategies (Donchian breakout, ATR-trailed momentum, Bollinger mean reversion, RSI mean reversion, moving-average crossover), ATR-based and Kelly-fraction position sizing, portfolio and operation views, and placing and canceling orders. Live trading is off by default: place_order requires both the TINKOFF_LIVE_TRADING=YES_I_KNOW environment variable and the dry_run=false argument at the same time; without both it returns a simulated response with the hypothetical order details.
Who it is for. For traders who want to backtest strategies and analyze the market with an agent, with real trading turned on deliberately and explicitly.
Good fit when
- You want to scan the market and find stocks in a given regime with specific indicators
- You want to backtest a ready strategy before trying it live
- You need ATR-based or Kelly-fraction position risk sizing before a trade
Not a fit when
- You are not ready to explicitly enable live trading through the dual condition: by default place_order only simulates the order
- You need a profitability guarantee: the author states plainly this is not investment advice and backtests do not predict future returns
- You are not prepared for the risk of total capital loss: the README warns about this directly in the disclaimer
Example request
Scan the market, find the top 3 stocks in a trending regime with RSI under 60, and backtest momentum_atr on each over 2 yearsLimitations
The license in GitHub's metadata is marked Other, though the README text references MIT, so check the LICENSE file directly. The author states plainly the project is for educational and research purposes and the authors are not registered investment advisors. Live trading needs explicit dual confirmation, but once enabled the server places real orders on the exchange.
How to disable. Remove the tinkoff entry from your MCP client config and make sure TINKOFF_LIVE_TRADING is not set anywhere in your environment.
MCP
- Transport
- stdio
- Authentication
- API key
| Environment variables | |
|---|---|
| TINKOFF_TOKEN required, secret | T-Invest API token; a read-only token is enough for analysis |
| TINKOFF_ACCOUNT_ID required | Account ID, needed for portfolio and orders |
| TINKOFF_MODE | real or sandbox |
| TINKOFF_LIVE_TRADING | Only the value YES_I_KNOW together with dry_run=false enables real orders |
Security check
- Once TINKOFF_LIVE_TRADING is explicitly enabled, the server places and cancels real exchange orders with account money
- Start with TINKOFF_MODE=sandbox and a read-only token, and enable live trading deliberately, one trade at a time
README in short
The README describes installing via pip, environment variables including the dual condition for live trading, connecting to Hermes Agent and Claude Desktop, a table of 12 tools, five built-in backtest strategies, a safety model with two conditions and a 1-lot order minimum, example prompts in Russian, and an explicit disclaimer about the project's educational nature and the risks of trading the Russian stock market.
FAQ
Could the agent accidentally buy or sell shares?
No, that needs both the TINKOFF_LIVE_TRADING=YES_I_KNOW variable and dry_run=false in the call itself; without both it only returns a simulation.
Can I test without a real account?
Yes, TINKOFF_MODE=sandbox switches the server into T-Bank Invest's sandbox.
Related
A set of investment research skills for Claude Code and Codex built on four investors' methodologies. It is a research tool, not investment advice
Stripe's official MCP server, plugins and skills: the agent searches the API and docs, reads and changes account data
Alpaca's official MCP server: orders for stocks, ETFs, crypto and options, position management and market data in plain language
inn-check-ru: counterparty risk traffic light
inn-check-ru
An open-source skill that checks a Russian company by tax ID across EGRUL, bailiffs, courts, bankruptcy and sanctions, and returns a dated risk verdict