DeepCode - HKUDS (The University of Hong Kong Data Science lab)

DeepCode

HKUDS (The University of Hong Kong Data Science lab) · AI-Native IDEs & Editors · Updated Sep 1, 2026

DeepCode is an open-source, general-purpose autonomous coding agent designed for real-world software engineering that handles tasks from understanding and implementation to verification. It features a persistent, goal-driven loop architecture that allows the agent to continuously work, test, and repair code until a task is verified.

Recent Updates

Released v2.1.0 on 2026-08-12, which unified Skills, Agent Plugins, and MCP (Model Context Protocol) across interfaces, following the major v2.0.0 release in August 2026 that introduced the Loop Engineering framework.

As of Sep 1, 2026

Key Features

Loop Engineering for autonomous implementation, verification, and repair
Verified delivery with execution evidence (test results, builds, and Diffs)
Durable, persistent Sessions with project-level memory and context
Unified interface across CLI and Tauri-based Desktop workbench
Support for custom Skills, automations, and multiple LLM providers

Who It's For

✓ Best For

Complex, multi-step software engineering tasks, repository maintenance, automated regression checks, and scientific code reproduction.

✗ Not Ideal For

Simple, one-shot code snippets or environments where users prefer native IDE extensions without terminal-first workflows.

Pricing

Free

Open-source software; users provide their own API keys for model providers (e.g., OpenRouter, OpenAI, Anthropic, DeepSeek, Gemini, Ollama, vLLM).

Verdict

A highly capable, open-source agentic tool that excels at long-horizon software engineering by prioritizing verified, durable results over simple code generation.

Track your AI stack