crypto-claude-desk

Crypto Trading Desk exposes 7 MCP servers that collectively provide 83 distinct tools for live crypto data, trading signals, backtesting indicators, risk assessment, and cognitive learning. You can interact with the system via slash commands (e.g., /crypto-trading-desk:quick BTC, /crypto-trading-desk:analyze ETH) or by natural language prompts that route to specialized agents. The data/analysis pipeline runs across dedicated servers: data gathering (crypto-data), market interaction (crypto-exchange), technical analysis (crypto-technical), risk and portfolio logic (crypto-futures, crypto-market-microstructure), advanced indicators (crypto-advanced-indicators), and persistent learning/storage (crypto-learning-db). When you request a full analysis, multiple agents coordinate in phased workstreams to generate reports, perform risk assessment, and propose executable trading decisions, all while writing reports to disk for traceability. Tools span market quotes, OHLCV data, indicators like RSI/MACD, sentiment/news, orderbook depth, funding rates, and learning-pattern tracking. You can also extend the system by creating new components with the system-builder workflow and test suites. The setup process ensures dependencies are present, MCP servers are responsive, and the environment is ready for ongoing trading analysis and learning.

Prerequisites:

• Git installed
• An environment with Python support and uv (or a compatible environment manager) if you plan to run MCP servers locally
• Network access to download dependencies and, optionally, to reach public APIs used by the MCP components

Install and run (recommended flow):

1. Clone the repository that contains the MCP server suite:

git clone https://github.com/hugoguerrap/crypto-trading-desk.git
cd crypto-trading-desk

2. Install dependencies and set up the environment. If using uv (Python), ensure it is installed or install Python dependencies required by the MCP servers:

# Example for Python/uv setup (adjust as needed for your environment)
python -m pip install --upgrade pip
python -m pip install uv

3. Start or register the MCP servers. The project auto-discovers agents and MCP servers when run via Claude Code. If running locally for development, you can start the environment and then trigger the MCP setup:

# Start the environment and let it discover MCP servers
# This may be done through your CLI workflow or via a startup script provided by the project
/crypto-trading-desk:setup

4. Verify that all 7 MCP servers are healthy. The setup step reports status and confirms each server is responsive. You should see a summary that all servers are running and ready to handle tool calls.

5. Run quick tests or demos:

/crypto-trading-desk:quick BTC

This should return a live market snapshot including price, volume, sentiment, and other metrics in a short time.

Notes:

• The exact runtime commands may vary depending on how you choose to run Claude Code with the plugin. The README’s setup flow is typical: install dependencies, ensure uv is present, verify MCP servers, and then run quick checks.
• If you prefer a containerized approach, you can adapt the steps to build/run a docker image for the entire stack once you have a suitable image name.

Tips and common considerations:

• The 7 MCP servers power 83 tools across data collection, market access, technical analysis, risk, learning, and system construction. Each server exposes focused capabilities (e.g., crypto-data for market metrics, crypto-technical for indicators, crypto-learning-db for persistence and learning tracks).
• No API keys are required for core capabilities, as described in the project: it leverages public APIs and local computations; changes in external data sources may affect results, so rely on the reports and learning logs for long-term performance tracking.
• When running experiments, use the learning database to track predictions, outcomes, and NL evaluations to build a robust pattern library and to improve future decisions.
• If you encounter environment issues (e.g., uv not found), ensure your OS has the required dependencies and permissions; the setup will attempt to install uv automatically where possible.
• If you add new MCP components or servers, follow the same discovery pattern so Claude Code can recognize and route to the new tools seamlessly.