wangziqi06/724-office

7/24 Office — Self-evolving AI Agent system. 36 tools, 10,000 lines pure Python, modular architecture, MCP plugins, three-layer memory, nudge system, AI mirror, 24/7 production.

What it solves

7/24 Office is a production-ready AI agent system designed to run 24/7 as a personal assistant. It solves the complexity of building reliable agents by avoiding heavy frameworks (like LangChain or LlamaIndex) in favor of a lightweight, modular Python architecture that focuses on durability, token efficiency, and self-maintenance.

How it works

The system operates as a multi-tenant router that provisions individual Docker containers for users, connecting them to a messaging platform via an HTTP server. At its core is a tool-use loop that handles OpenAI-compatible function calling with automatic retries. It manages context through a three-layer memory system: short-term session history, LLM-compressed long-term facts, and a LanceDB vector database for active recall. To ensure reliability, it employs a "nudge system" for behavioral correction and a circuit breaker to disable failing tools.

Who it’s for

It is designed for developers who want a transparent, framework-free agent implementation that can be deployed on edge hardware (like Jetson Orin Nano) and integrated into messaging platforms for personal or group use.

Highlights

  • Zero Framework Dependency: Built with pure Python to ensure every line is debuggable and transparent.
  • Self-Evolving Capabilities: The agent can write, save, and load new Python tools at runtime using the create_tool function.
  • Three-Layer Memory: Combines session history, compressed structured facts, and vector retrieval to optimize token usage.
  • Extensive Toolset: Includes 36 built-in tools for web search, video processing (trimming/BGM), ECharts visualization, and scheduling.
  • MCP Integration: Supports the Model Context Protocol (MCP) to connect external servers via JSON-RPC.
  • Self-Repair: Features daily self-checks, session health diagnostics, and automated error log analysis.

Related

  • Project
    fuyuxiang/echo-agent

    A self-hosted, long-running AI agent featuring a four-layer cognitive memory system and a self-evolution engine to improve skills from real-world usage.

  • Project
    luoguoxiong/aipack

    A lightweight personal AI assistant framework featuring a custom Agent runtime, hybrid long-term memory, and multi-level context compression.

  • Project
    MervinPraison/PraisonAI

    PraisonAI is an open‑source framework for building autonomous, self‑improving AI agents. It implements a five‑layer stack (prompt, context, harness, loop, graph) plus a managed‑agents layer that lets you run agents locally or in remote sandboxes (Docker, e2b, Modal, etc.). The Python SDK (`praisonaiagents`) provides declarative agents, tool integration via the MCP protocol, loop‑control (budgets, doom‑loop detection), and graph orchestration (routing, parallel, repeat). A CLI, visual flow builder, dashboard, and JavaScript SDK round out the ecosystem. Supports 100+ LLM providers and is MIT‑licensed.

  • Project
    zeenie-ai/OpenCompany

    An open-source, self-hosted no-code canvas for building AI agent workflows that connects LLMs to over 30 services and allows for local model execution.

  • Project
    TencentCloud/TencentDB-Agent-Memory

    A shared memory server and management hub that allows AI agents to persist and share experience, skills, and project knowledge across different sessions and frameworks.