Skip to content
Conduit

mcp server

cargo-agent

Self-evolving AI Agent CLI in Rust. 60+ tools for coding, testing, refactoring, trading. Dual-licensed: MIT + Commercial.

kaizhong8888-rgb/cargo-agentcommunitylistedunreviewed

Set up

Runs on your computer. Pick your agent and operating system.

This project does not publish an install method we can turn into a config. Follow the README instead.

README

synced Oct 23, 2026

cargo-agent

Self-evolving AI Agent CLI written in Rust. 60+ built-in tools for coding, analysis, refactoring, trading, and more.

What is cargo-agent?

cargo-agent is an autonomous AI programming assistant that runs as a CLI tool. It connects to LLM APIs (OpenAI, Anthropic, Ollama) and provides a tool-use loop where the agent can read/write files, execute code, run git operations, analyze code structure, refactor, test, and even modify its own source code.

Key Features

🤖 AI Agent Core

  • Tool-use loop: 60+ built-in tools the agent can autonomously call
  • Multi-model support: OpenAI, Anthropic, Ollama (auto-adaptation)
  • Model router: Automatic model selection based on task complexity
  • Streaming responses: Real-time output with token usage display

🔧 60+ Built-in Tools

CategoryTools
File/FSRead, write, list, grep, diff, archive
Code AnalysisAST analyzer, code quality, code review, Clippy lint
Code ExecutionSandboxed cargo run/build/test/check/clippy
RefactoringSmart refactor: derive, rename, unwrap, visibility
TestingTest generator, fuzz driver, benchmark
GitStatus, diff, log, clone, commit, push, workflow
DependenciesAdd, remove, update, tree, audit, outdated
MemorySQLite-backed persistent memory with TF-IDF search
TasksTask planner, task pool, TODO manager
LLMCall external LLMs for code gen/review
DatabaseSQL queries, table management, migrations
CryptoAES encrypt/decrypt, hash, sign, JWT
TradingQuantitative backtesting, strategies, indicators
SecuritySecurity scanner, license audit
CI/CDCI/CD pipeline tools
DataCSV/JSON processing, chart generation
NetworkHTTP client, GitHub API, OpenAPI spec generation
CommunicationEmail, webhooks, templates, PDF generation
SystemSystem monitor, Docker generation, cross-compile

🧬 Self-Evolution System

  • Source code modification: Safely read, write, create, delete files
  • Tool creation: Dynamically generate new tools and integrate them
  • Verification: Auto-run cargo check after modifications
  • Memory persistence: Durable knowledge with tags and importance levels
  • Evolution tracking: Record tool creation, code changes, lessons learned
  • Self-reflection: Analyze past behavior, identify improvement patterns

📊 Quantitative Trading Module

  • Backtesting engine: Position sizing, stop-loss, walk-forward validation
  • 6+ strategies: SMA crossover, RSI mean-reversion, MACD, Bollinger Bands, Triple EMA, VWAP-RSI
  • 26 submodules: ML models, factor analysis, feature engineering, portfolio optimization
  • Market data: Fetch candle data, Chinese stock market integration
  • Risk management: Market regime detection, portfolio optimizer

💬 MCP Server

  • Model Context Protocol: Expose all 60+ tools to MCP-compatible clients
  • JSON-RPC 2.0: Standard protocol over stdio
  • External integration: Connect with any MCP-compatible IDE or agent

Quick Start

Prerequisites

  • Rust 1.75+ (edition 2021)
  • An LLM API key (OpenAI, Anthropic, or Ollama for local)

Install

# From crates.io (coming soon)
cargo install cargo-agent

# From source
git clone https://github.com/kaizhong8888-rgb/cargo-agent.git
cd cargo-agent
cargo build --release

Configure

Config lives at ~/.cargo-agent/config.yaml:

model:
  api_key: "your-api-key"
  base_url: "https://api.openai.com/v1"  # OpenAI-compatible
  model: "gpt-4o"

API key resolution order: config file → CARGO_API_KEY → OPENAI_API_KEY → ANTHROPIC_API_KEY

Usage

# Interactive REPL mode
cargo run

# One-shot mode (bypasses UI, outputs response directly)
cargo run -- run "Analyze the code quality of src/"

# Run with specific prompt
cargo run -- run "Create a new CLI project with serde and tokio"

Example Prompts

→ 分析 src/ 目录的代码质量
→ 为 lib.rs 生成单元测试
→ 创建一个新的 CLI 项目
→ 运行 clippy 并修复常见问题
→ 回测 SMA 交叉策略在 A 股的表现
→ 搜索 crates.io 上最好的 Rust web 框架

Architecture

┌─────────────────────────────────────────────────┐
│                  User Interface                  │
│         REPL Loop / One-shot / MCP / TUI         │
├─────────────────────────────────────────────────┤
│                   Gateway                        │
│     ModelClient → ToolRegistry → SkillRegistry  │
├─────────────────────────────────────────────────┤
│                   AIAgent                        │
│  Chat loop → Tool execution → Memory injection  │
├─────────────────────────────────────────────────┤
│                  Tool Layer                      │
│    60+ Tools: File, Code, Git, DB, Trading...   │
├─────────────────────────────────────────────────┤
│                Infrastructure                    │
│    LLM APIs | SQLite | Git | FS | Network       │
└─────────────────────────────────────────────────┘

Project Structure

src/
├── main.rs              # CLI entry point
├── lib.rs               # Public module exports
├── gateway/             # Orchestrator
├── agent/               # AIAgent chat loop
├── model/               # LLM API client + router
├── tools/               # 60+ builtin tools
│   └── builtin/
├── trading/             # Quantitative trading (26 modules)
├── memory/              # SQLite memory store
├── skills/              # YAML skill system
├── config/              # YAML configuration
├── ui/                  # Terminal UI
├── tui/                 # Full-screen dashboard
└── mcp/                 # MCP server

💼 Need a Custom AI Agent?

I build enterprise-grade AI agents. cargo-agent is my portfolio — 48K+ lines of Rust, 60+ tools, MCP server.

ServiceStarting From
Technical Consulting (2hr)¥1,500
Agent Architecture Design¥8,000+
Custom Agent Development¥30,000+
Enterprise Integration¥50,000+

Free 30-min consultation. Open an issue or see SERVICES.md.

License

Dual-licensed:

  • MIT License — Free for personal and open-source use. See LICENSE for details.
  • Commercial License — For enterprise use, contact us for pricing and support.

Commercial Use

For commercial licensing, enterprise support, or custom AI Agent development services, see PRICING.md.

Contributing

We welcome contributions! See CONTRIBUTING.md for our test coverage standards.

Links

↑↓ navigate openesc close