deeppowers

DeepPowers is a Fully Homomorphic Encryption (FHE) based collaboration framework tightly integrated with the MCP (Model Context Protocol) ecosystem. It enables privacy-preserving computation across MCP workflows by performing operations on encrypted data, ensuring that plaintext data never leaves client or trusted environments. The server component handles secure storage, task dispatch, and scheduling for encrypted tasks, while leveraging a Trusted Execution Environment (TEE) for secure ciphertext processing. This setup supports interoperability with multiple LLM providers and MCP servers, enabling confidential execution of prompts, tooling, and data processing. To use it, run the server component to participate in MCP task submission, encrypted computation, and verification flows, and rely on the integrated privacy-preserving primitives to maintain data confidentiality throughout the lifecycle of a task.

Key capabilities include end-to-end encryption for MCP interactions, secure task execution within TEEs, seamless MCP integration for privacy-preserving workflows, and cross-model compatibility with providers like DeepSeek, GPT, Gemini, and Claude. The framework emphasizes reduced data leakage risk, lower latency through optimized encrypted computation, and auditability via blockchain or distributed ledger techniques for encrypted results and proofs. When used in an MCP ecosystem, DeepPowers acts as a privacy-first compute node, enabling secure evaluation of models and tools while preserving user data privacy.

Prerequisites:

• Python 3.8 or newer
• Git
• (Optional) Python virtual environment tools (venv, virtualenv)

1. Clone the repository or install via PyPI (if available):

   • git clone https://github.com/deeppowers/deeppowers
   • cd deeppowers

2. Create and activate a virtual environment:

   • python -m venv venv
   • source venv/bin/activate # On Windows use: venv\Scripts\activate

3. Install dependencies:

   • pip install -r requirements.txt
   • If there is a backend/TEE specific dependency, install as indicated in docs (e.g., git submodules or system packages).

4. Configure environment variables (example):

   • export DEEPPPOWERS_CONFIG=path/to/config.yaml
   • export FHE_BACKEND=OpenFHE
   • export LOGGER_LEVEL=INFO

5. Run the server:

   • python -m deeppowers

6. (Optional) Run validation or tests if provided:

   • python -m pytest tests

Notes:

• Ensure that any required native libraries (for FHE backends or TEE) are installed per the backend documentation.
• If you install via pip, adjust the commands accordingly (e.g., pip install deeppowers if published on PyPI).

Tips and common issues:

• Ensure your FHE backend and TEE environment are correctly configured before starting the server; misconfigurations can lead to encrypted computation failures.
• If tasks hang or exceed latency budgets, check the MEC (multi-node) synchronization and ensure consensus nodes are reachable.
• Use the provided env vars to tune verbosity (LOGGER_LEVEL) and to point to the correct configuration file (DEEPPPOWERS_CONFIG).
• For debugging, enable verbose logging and monitor task queues and node statuses through the MCP tooling or the server's admin endpoints.
• If upgrading, review compatibility notes between MCP versions and the DeepPowers backend integrations to avoid protocol mismatches.