deeppowers

MCPAI代理

提供方 deeppowers · 包名 deeppowers

AI代理 stdio · 本地进程 未标注 社区 / 第三方 Claude DesktopCursorClineClaude Code
查看接入方式

DEEPPOWERS is a Fully Homomorphic Encryption (FHE) framework built for MCP (Model Context Protocol), aiming to provide end-to-end privacy protection and high-efficiency computation for the upstream an

一句话选型结论

按资料字段整理

deeppowers 属于AI代理类 MCP 服务器,stdio · 本地进程,鉴权方式未标注,可直接接入 Claude Desktop、Cursor、Cline 等,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类AI代理 传输方式stdio · 本地进程 鉴权未标注 部署方式本地部署
GitHub 星标 276 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

AI代理
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

accelerator ai llm-inference llms

支持客户端

适用场景

acceleratoraillm-inference

安装与配置

stdio · 本地进程 未标注
客户端在本机拉起子进程,数据不出本机
git clone https://github.com/deeppowers/deeppowers && cd deeppowers

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Cross-platform,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
给代理接外部能力

让代理在任务中途调用搜索、数据库等真实工具。

多步任务编排

把一串操作拆成步骤,由代理按顺序执行。

自动化重复流程

把固定流程交给代理跑,减少手工操作。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Cursor、Cline,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在AI代理场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Cursor 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
deeppowers当前 未标注 stdio · 本地进程 本地部署 有 2026-04-19
Filesystem 560c4363 13 未标注 本地部署 无 2026-05-17
Server Everything (everything) 11 stdio · 本地进程 本地部署 有 2026-05-17
Time 7365cb6d 2 SSE · 服务端推送 远程部署 有 2026-05-18
Everything Claude Code 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Openmemory 557e1585 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

AI代理
字段来自项目仓库与公开索引,中文说明由本站标注
提供方deeppowers上游标注的维护方 deeppowers
分类AI代理本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式未标注远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时deeppowers · Cross-platform本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2026-04-19上游最后核验时间

详细介绍

13.7 KB
来自项目仓库与上游目录的原文,未做删改

DeepPowers

DEEPPOWERS is a Fully Homomorphic Encryption (FHE) collaboration framework built for MCP (Model Context Protocol), aiming to provide end-to-end privacy protection and highly efficient computation for the upstream and downstream ecosystems of the MCP protocol. By deeply integrating the FHE framework with the MCP protocol, we are dedicated to creating a secure, efficient, and scalable computing framework for MCP. This ensures that data remains encrypted throughout transmission, storage, and computation, while also supporting complex computing logic, eliminating unnecessary data transmission and computation, and thereby rapidly improving MCP's operational efficiency.

By removing delays in MCP interactions, it aims to provide robust momentum for the Model Context Protocol (MCP) ecosystem. DEEPPOWERS unleashes a higher level of efficiency, collaboration, and performance for MCP workflows. Support for various MCP servers and leading large language models (LLMs) such as DeepSeek, GPT, Gemini, and Claude ensures unparalleled versatility and enhanced collaborative efficiency.

Overview

Fully Homomorphic Encryption (FHE) allows computations (such as addition, multiplication, etc.) to be performed directly on encrypted data without decryption. The computation results remain encrypted, and only authorized users can decrypt them. FHE resolves the conflict between data privacy and computational efficiency and is suitable for scenarios such as cloud computing, medical data analysis, and financial transactions. Its core lies in ensuring data remains encrypted throughout the process, eliminating the risk of privacy leakage in intermediate steps, while supporting complex computations, providing the ultimate guarantee for data security and compliance. It supports languages such as C++, Python, and CUDA, facilitating integration into the existing MCP ecosystem.

Key features

  • End-to-end Encryption: Complete protection for user-LLM interactions with data remaining encrypted throughout the entire process
  • Secure Task Execution: Encrypted data transmission and computation for sensitive operations without exposure to third parties
  • Seamless MCP Integration: Deep integration with the Model Context Protocol ecosystem, enhancing MCP's capabilities with privacy-preserving computations
  • Accelerated MCP Workflows: Process encrypted data locally and reduce unnecessary transfers, significantly improving MCP operational efficiency
  • Enhanced Tool Security: Enable secure execution of MCP tools that handle sensitive data, expanding the scope of privacy-preserving tasks
  • Cross-Model Compatibility: Work with various MCP servers and LLM providers including DeepSeek, GPT, Gemini, and Claude
  • Reduced Latency: Minimize delays in MCP interactions by optimizing the encryption, computation, and decryption processes
  • Privacy-Preserving AI: Complete AI assistance workflows while maintaining strict data privacy throughout
  • Scalable Architecture: Designed to grow with the MCP ecosystem, supporting emerging models and computational tools

Powered by Concrete-ML

DeepPowers is built on Concrete-ML, an open-source library developed by Zama for privacy-preserving machine learning using Fully Homomorphic Encryption.

Concrete-ML provides:
- Privacy-preserving ML framework using FHE
- Compatibility with traditional ML frameworks (scikit-learn, PyTorch)
- Tools for quantizing and converting ML models for FHE execution
- Simple and intuitive APIs similar to scikit-learn

With Concrete-ML as its foundation, DeepPowers extends these capabilities to LLM interactions, creating a secure environment for AI-assisted tasks.

Structure

1. MCP Client

  • Responsibilities:
  • Local FHE encryption/decryption
  • Key pair generation and management (private key never leaves the client)
  • Task submission and encrypted result display
  • User authentication and permission control

2. MCP Server

  • Responsibilities:
  • Secure storage and dispatching of encrypted data and tasks
  • Task management and scheduling
  • Node status monitoring and logging
  • No access to plaintext data

3. TEE Compute Node

  • Responsibilities:
  • Executes encrypted computations using FHE libraries within a Trusted Execution Environment (TEE)
  • Ensures data and computation security during runtime
  • Outputs encrypted results and computation proofs

4. Decentralised Review Nodes

  • Responsibilities:
  • Perform consensus and verification of computation results and proofs
  • Ensure consistency and integrity of results
  • Write audit logs and consensus records to the blockchain for traceability

5. Blockchain

  • Responsibilities:
  • Store audit logs, computation proofs, and operation records
  • Ensure full traceability and immutability of the process
  • Facilitate compliance with regulations (e.g., GDPR, HIPAA)

To experience the complete workflow, check out our Demo.

Key Technology Integration

FHE Encryption and Ciphertext Computation

  • Encryption Algorithm Selection: Utilizes industry-leading FHE libraries (such as Microsoft SEAL, OpenFHE, IBM HELib), supporting encryption and computation for various data types including integers and floating-point numbers.
  • Ciphertext Computation Process: All AI inference, data analysis, and statistical tasks are completed in the encrypted state, with results remaining encrypted.
  • Key Management: Users generate and maintain FHE key pairs locally, ensuring private keys never leave the local environment to maximize protection against key leakage.

Trusted Execution Environment (TEE)

  • TEE Function: Provides trusted execution guarantees for performance bottleneck processes (such as ciphertext batch processing, key exchange), preventing malicious access to runtime memory and registers.
  • FHE and TEE Integration: Loading FHE libraries within TEE further enhances the security and trustworthiness of ciphertext computation.

Decentralized Verification and Blockchain Consensus

  • Multi-Node Consensus: Through blockchain or distributed ledger technology, multiple verification nodes perform consistency checks and reach consensus on ciphertext computation results, ensuring they cannot be tampered with.
  • Transparency and Traceability: Every operation is traceable, enhancing process transparency and user trust.

Performance Optimization and Engineering Implementation

  • Batch Processing and Pipeline Optimization: Optimizing ciphertext batch operations and pipeline processing to reduce FHE computation latency.
  • Algorithm Adaptation and Simplification: Adapting AI models and statistical algorithms for FHE characteristics, reducing multiplication depth and optimizing operation gate counts.
  • Frontend Integration: Integrating FHE encryption/decryption functions into terminal applications through WebAssembly or Native SDK to enhance user experience.

Development

As part of our development effort, we are implementing significant hardware optimizations for Concrete-ML to accelerate encrypted computation processes. These optimizations are crucial for improving the performance of homomorphic encryption operations, which are typically computation-intensive.

Our hardware optimization initiatives include:
- Custom hardware acceleration for key FHE operations
- Parallel processing implementations for encrypted data
- Low-level optimizations for cryptographic primitives
- Specialized memory management for FHE workflows

These optimizations will enable DeepPowers to process encrypted data more efficiently, making privacy-preserving AI interactions faster and more practical for real-world applications.

Technical Challenges and Solutions

1. High Performance Overhead of FHE Computation

  • Challenge: FHE operations are fundamentally several orders of magnitude slower than plaintext computation, especially with large-scale data or complex models, resulting in high latency and resource consumption.
  • Solutions:
  • Utilize the latest FHE libraries with SIMD and other batch processing technologies to improve throughput.
  • Prioritize supporting low-depth, low-complexity AI/data analysis tasks for practical application scenarios.
  • Combine with TEE, using secure hardware acceleration for non-core computations to balance security and performance.

2. Key Management and User Experience

  • Challenge: FHE keys, once lost, make data recovery impossible, making local key management and backup critical.
  • Solutions:
  • Provide multiple key management methods including local secure storage, hardware security modules (HSM), or mnemonic phrases.
  • Design user-friendly key recovery mechanisms to lower operational barriers.

3. Algorithm Compatibility and Model Adaptation

  • Challenge: FHE currently has limited support for certain non-linear operations (such as ReLU, Sigmoid), making complex AI models difficult to migrate directly.
  • Solutions:
  • Prioritize FHE-friendly algorithms such as linear models, decision trees, and logistic regression.
  • Research homomorphic approximation methods, replacing complex operations with polynomial approximations.
  • Continuously track advances in the FHE field to dynamically upgrade platform capabilities.

4. Efficiency and Security of Decentralized Verification

  • Challenge: Multi-node consensus introduces latency and network load, with the risk of malicious nodes persisting.
  • Solutions:
  • Adopt efficient Byzantine Fault Tolerance (BFT) consensus algorithms to improve verification efficiency.
  • Introduce reputation mechanisms and economic penalties to incentivize honest participation.

5. End-to-End Security and Compliance

  • Challenge: Proving the platform does not leak user data throughout the process while meeting strict regulations like GDPR and HIPAA.
  • Solutions:
  • Provide complete security audit logs and zero-knowledge proofs to demonstrate no plaintext data transfer on the platform.
  • Regularly undergo third-party security audits and penetration testing.

Technical Achievements

  1. End-to-End Encrypted Data Collaborative Computation
    User data remains encrypted throughout collection, upload, computation, verification, and return, preventing the platform or third parties from seeing plaintext.

  2. Integration of FHE with TEE, Blockchain, and Other Cutting-Edge Technologies
    Breaking single technology barriers to achieve an innovative architecture with multiple security guarantees.

  3. Enabling Compliant AI/Data Analysis Deployment in Highly Sensitive Industries
    Allowing medical, financial, and other sectors to fully extract data value without compromising privacy.

  4. Providing a Replicable Technical Paradigm for Distributed Computing and Privacy Protection
    Establishing a modular, scalable system architecture for future technology upgrades and business expansion.

Application Scenarios

  • Medical Joint Modeling: Multiple hospitals can jointly train AI models to improve diagnostic accuracy without sharing original patient cases, protecting patient privacy.

  • Financial Risk Control Collaboration: Banks and insurance companies can jointly analyze fraud risks with fully encrypted customer information, preventing sensitive data leakage.

  • Cross-Department Government Data Analysis: Tax, public security, social security, and other departments can jointly analyze population movement and economic trends without concerns about data misuse.

  • Cloud AI Inference Services: Users can upload encrypted sensitive data, with cloud AI models inferring directly on ciphertext, preventing cloud service providers from accessing inputs or results.

  • Privacy-Protecting Data Market: Individuals or enterprises can securely circulate and monetize encrypted data with dual guarantees of data ownership and privacy.

Roadmap

The DeepPowers SDK development roadmap includes:

Phase 1: Core Infrastructure

  • [x] FHE integration with MCP framework
  • [x] Basic encryption protocols for LLM communication
  • [x] Prototype SDK for developer testing
  • [x] Initial hardware acceleration implementations

Phase 2: Feature Expansion

  • [ ] Enhanced model support for various LLM architectures
  • [ ] Expanded API capabilities for diverse applications
  • [ ] Performance optimizations for real-time interactions
  • [ ] Advanced hardware optimizations for Concrete-ML
  • [ ] GPU/FPGA acceleration for critical FHE operations

Phase 3: Enterprise Integration

  • [ ] Advanced security features for organizational deployment
  • [ ] Compliance and audit capabilities
  • [ ] Scalable infrastructure for high-volume applications
  • [ ] Fully optimized hardware acceleration with specialized processing units
  • [ ] Enterprise-grade performance for encrypted computations

Coming Soon

The DeepPowers SDK is currently under active development. Our team is working diligently to bring this technology to developers and organizations who prioritize privacy in their AI implementations.

We plan to introduce a native token-based incentive mechanism for the DeepPowers system, rewarding ecosystem participants such as compute nodes, verification nodes, and data contributors based on task completion and service quality, while implementing staking and penalty mechanisms to ensure network security and trustworthiness. The token will not only serve as an incentive but also empower platform governance, service payments, and ecosystem collaboration, fostering community engagement and sustainable growth. Moving forward, this token incentive system will be a key driver for the prosperity of the DeepPowers ecosystem and the practical adoption of privacy-preserving computation.

Stay tuned for updates on our progress and early access opportunities.

For inquiries about DeepPowers, please contact our development team.

FHEDEEP CA:6uHQEuSFwrCYKByHfkVgWrGWTESRZpNm951PaqxWpump

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
deeppowers 怎么安装?

在客户端按下面的安装命令接入:git clone https://github.com/deeppowers/deeppowers && cd deeppowers。对应的包名是 deeppowers。装完在客户端配置里引用同一个包或端点。

deeppowers 怎么在 Claude Desktop、Cursor、Cline 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

deeppowers 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

deeppowers 需要 API Key 吗?

资料没有标注鉴权方式,是否需要密钥取决于服务本身,接入前按官方文档确认。

deeppowers 支持哪些客户端?

资料标注支持 Claude Desktop、Cursor、Cline、Claude Code;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

deeppowers 能提供哪些能力?

资料显示能力标签包括 accelerator、ai、llm-inference、llms。具体工具清单以官方文档为准。

deeppowers 需要什么运行环境?

运行时为 Cross-platform;属于本地进程,需要在本机装好运行环境。

deeppowers 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-04-19。