What do I need to run Composer MCP Server?
You need Python and the package manager `uv`. Installation via DXT or manual config is quick; no account is needed for basic backtesting.
@invest-composer
Basic information
Composer's MCP server lets MCP-enabled LLMs like Claude backtest trading ideas and automatically invest in them for you
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"composer": {
"command": "uvx",
"args": [
"composer-trade-mcp"
],
"env": {
"COMPOSER_API_KEY": "<insert-your-api-key-here>",
"COMPOSER_SECRET_KEY": "<insert-your-api-secret-here>"
}
}
}
}Define an automated strategy using Composer's system.
Backtest a symphony that was created with `create_symphony`
Backtest an existing symphony given its ID
Save a symphony to the user's account
Update a saved symphony
List all brokerage accounts available to the Composer user
Get the holdings of a brokerage account
Get the aggregate portfolio statistics of a brokerage account
Get stats for every symphony in a brokerage account
Get daily performance for a specific symphony in a brokerage account
Get the daily performance for a brokerage account
Get the definition about an existing symphony given its ID.
Get market hours for the next week
Invest in a symphony for a specific account
Withdraw money from a symphony for a specific account
Cancel an invest or withdraw request that has not been processed yet
Skip automated rebalance for a symphony in a specific account
Immediately sell all assets in a symphony
Immediately sell all assets in a symphony (or queue for market open if outside of market hours)
Perform a dry run of rebalancing across all accounts to see what trades would be recommended
Perform a dry run of rebalancing for a specific symphony to see what trades would be recommended
Rebalance a symphony NOW instead of waiting for the next automated rebalance
Execute a single order for a specific symbol like you would in a traditional brokerage account
Cancel a request for a single trade that has not executed yet
Composer MCP Server is the official Model Context Protocol server from Composer that lets MCP-enabled LLMs (Claude Desktop, Cursor, OpenAI Agents, etc.) validate investment ideas through backtests and trade multiple automated strategies—called “symphonies”—in parallel to compare live performance.
Install uv (Python package manager), then either download and drag the composer-trade-mcp.dxt file into Claude Desktop’s Extensions settings, or add a manual configuration in claude_desktop_config.json using uvx composer-trade-mcp. Backtesting works without a Composer account; full account features require an API key.
You need Python and the package manager `uv`. Installation via DXT or manual config is quick; no account is needed for basic backtesting.
An API key is required to interact with your Composer account (view holdings, save symphonies). Trading symphonies requires a paid subscription, but you can liquidate positions without one. A 14-day free trial is available.
It offers 26 tools including `create_symphony`, `backtest_symphony`, `save_symphony`, `get_account_holdings`, `get_portfolio_daily_performance`, `invest_in_symphony`, `withdraw_from_symphony`, `liquidate_symphony`, `execute_single_trade`, and more.
Tools that affect your funds: `invest_in_symphony`, `withdraw_from_symphony`, `skip_automated_rebalance_for_symphony`, `go_to_cash_for_symphony`, `liquidate_symphony`, `rebalance_symphony_now`, and `execute_single_trade`. Do not set these to “Always Allow”.
Logs are located at `%APPDATA%\Claude\logs\mcp-server-composer.log` on Windows and `~/Library/Logs/Claude/mcp-server-composer.log` on macOS.

Yuppsy MCP: build and publish real estate websites from your AI assistant. Yuppsy lets your AI assistant create live websites for real estate agents and individual property listings. You don't need a designer or any code
Hindsight: Agent Memory That Learns

Health by Dr Brad works out a health plan from your own results: which screening is due at your age, and evidence-based suggestions, each with the reason and the published guideline or study behind it. Your record is one
Open-source MCP to use Blender with any LLM
🚀 The fast, Pythonic way to build MCP servers and clients.
Comments