AideAgent: A Desktop AI Assistant Supporting Multi-Models and 28 Built-in Tools

AideAgent is a powerful desktop AI assistant developed and maintained by quanzefeng. The project is open-sourced on GitHub (link: https://github.com/quanzefeng/AideAgent), released in 2025, and continuously updated (latest update date: 2026-06-05). Key features include:

• Multi-model support: Compatible with mainstream large language models such as DeepSeek, GLM, Qwen, Claude, and MiniMax, while supporting local deployment (e.g., llama.cpp, LM Studio);
• 28 built-in tools: Covering areas like file operations, code collaboration, web search, Git management, and knowledge base retrieval;
• Advanced features: RAG knowledge base, MCP extension protocol, cross-session persistent memory, and automatic compression continuation function;
• Security design: Multi-layer protection ensures usability in production environments. This thread will introduce AideAgent's background, functions, features, and applicable scenarios in detail across different floors.

There are two major limitations in the current AI assistant field: most products are limited to web-based chat forms, or have single functions that cannot meet complex needs. AideAgent's design philosophy is to break this situation and become a desktop assistant that truly integrates into the daily workflow of developers and knowledge workers. It not only provides conversational AI services but also directly performs practical tasks such as code writing, file management, Git operations, and knowledge base queries, realizing the evolution from "chatbot" to "all-round assistant."

AideAgent adopts an open multi-model architecture, supporting presets from 8 model vendors with one-click switching:

• Cloud models: DeepSeek (V4-Flash/V4-Pro), GLM (GLM-4.7-Flash/GLM-4-Plus), Qwen (Qwen3.7-Max/Qwen-Plus), Claude (Sonnet4.6/Opus4.7/Haiku4.5), MiniMax (M2.7/M2.7-Highspeed);
• Local models: llama.cpp, LM Studio, Ollama;
• API compatibility: Supports OpenAI standards and Anthropic formats, allows custom API endpoints, balancing privacy protection and cost control needs.

AideAgent has 28 built-in practical tools covering the entire development workflow:

1. File and code operations: file_read/write/edit (precise text replacement), grep (regex search), glob (file lookup), lsp (code intelligence features like jump to definition);
2. Command and network: bash (cross-platform command execution with dangerous operation detection), web_search (Tavily priority + meta-search fallback), web_fetch (web scraping + SSRF protection);
3. Version control: git_diff/commit/branch (secure submission against injection), gh_pr/issue/repo (GitHub collaboration);
4. Knowledge management: kb_search/write (FTS5 + vector hybrid search), write_memory (cross-session memory);
5. Task collaboration: TaskCreate/Update/List (task tracking), skill system (custom workflow), Agent (sub-agent parallel tasks).

AideAgent's advanced features enhance its intelligence and expandability:

• RAG Knowledge Base: Hybrid search (FTS5 keyword + vector semantics + RRF fusion), supports MiniLM-L6 (offline) and Ollama custom models, automatically injects relevant notes into the conversation context;
• MCP Extension: Supports Anthropic's Model Context Protocol, connects to external tool servers (local process/remote HTTP), and unifies calls to built-in and external tools;
• Persistent Memory: Cross-session storage (multi-file Markdown with type tags like preferences, decisions), automatic deduplication and semantic selection, silently extracts long-term memory after conversations;
• Automatic Compression Continuation: Triggered when context tokens reach 90%, intelligently summarizes historical conversations, and seamlessly continues tasks (up to 250 rounds of effective work).

AideAgent excels in security and applicable scenarios: Security Design:

• Command execution: The bash tool includes dangerous operation detection and Hook interception;
• Git submission: Secure spawn mechanism to prevent injection;
• Network: web_fetch has built-in SSRF protection;
• Privacy: Knowledge base and memory are stored locally.

Applicable Scenarios:

1. Developers' daily work: Code writing, refactoring, debugging;
2. Knowledge management: Personal knowledge base construction and intelligent retrieval;
3. Project collaboration: Git operations, PR/Issue management;
4. Research and learning: Web search, data organization;
5. Automation: Implement complex workflows through skills and sub-agents.

AideAgent represents the development direction of desktop AI assistants: open (multi-model), capable (rich tools), intelligent (RAG + memory + self-evolution), and secure (multi-layer protection). It upgrades AI from a "conversational interface" to a "work partner," truly realizing the practical value of large language models.

In the future, with the expansion of the MCP ecosystem and the integration of more tools, AideAgent's capability boundary will continue to expand. For efficiency-seeking developers, researchers, and knowledge workers, this is a tool worth trying.