</>MCP Agents Market
Agent

DeepCode

by HKUDS16.4kPythonUpdated 2026-08-19

"DeepCode: Open Agentic Coding ((Agent Harness & Loop Engineering & Multi-Agent Orchestration)"

DeepCode DesktopDeepCode CLI

DeepCode is an open-source agentic coding framework that combines agent harness architecture, loop engineering, and multi-agent orchestration to transform research papers and natural language into production-ready code. The AI sub-agent system provides TUI, Desktop, and Web interfaces sharing a unified local service, supporting durable sessions, goal-driven development, evidence-based verification, and automated code generation workflows. DeepCode enables developers to tackle complex repository work through intelligent orchestration of specialized agents for understanding, planning, implementation, and verification. Originally designed for Paper2Code research reproduction, it now serves as a general-purpose coding agent framework with Skills, MCP integration, and automation capabilities.

Key Features

Multi-agent orchestration with specialized roles (Intent Understanding, Document Parsing, Code Planning, Code Generation, Reference Mining, Indexing, and Central Orchestrator)
Three unified interfaces (TUI, Desktop, Web) sharing durable sessions, models, Skills, permissions, and task state through a single local background service
Goal-driven Loop Engineering that continuously analyzes, implements, verifies, and repairs code around natural-language objectives
Evidence-driven completion with verification through tests, builds, diagnostics, diffs, and generated artifacts
Paper2Code workflow that processes research papers, technical documents, and reference repositories into executable implementations
Reusable Skills system compatible with DeepCode, Claude, and Agent Plugins 1.0 formats for team conventions and domain knowledge
Generic MCP server runtime supporting stdio, SSE, and Streamable HTTP with OAuth, timeouts, and session-owned lifecycle
Parallel agent execution in isolated Git worktrees with explicit conflict handling and file collision prevention

Use Cases

  • 01Reproducing research papers from academic publications into working code implementations with automated verification
  • 02Complex multi-file refactoring and feature development across large codebases with continuous verification loops
  • 03Automated repository maintenance including test repair, regression checks, and documentation upkeep via scheduled Automations
  • 04Code review and security analysis workflows using bundled upstream Skills from OpenAI, Codex, and Anthropic
  • 05Understanding unfamiliar codebases through read-only analysis sessions with project structure exploration
  • 06Frontend and web application development with design-to-implementation workflows and iterative testing

Related Agents

View more

hermes-agent

The agent that grows with you

Agent250.6k

agency-agents

A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.

Agent155.6k

openinterpreter

A coding agent for open models like Kimi K3

Agent68k

cline

Autonomous coding agent as an SDK, IDE extension, or CLI assistant.

Agent66.3k

DeepCode — FAQ

What is DeepCode and what does it do?+

DeepCode is an open-source agentic coding framework that uses multi-agent orchestration to understand projects, write code, run tests, and verify results. It can transform research papers and natural language goals into production-ready code through specialized agents for understanding, planning, implementation, and verification.

How do I install DeepCode?+

Install DeepCode with uv: run 'uv tool install --python 3.12 deepcode-hku' then 'deepcode init'. Requires Python 3.12 or later. For Desktop interface, download the native app from GitHub Releases. For source installation, clone the repo, build the web frontend with npm, and install with 'uv tool install --python 3.12 --force .'.

Which AI clients and models does DeepCode work with?+

DeepCode provides its own TUI, Desktop, and Web interfaces rather than plugging into external AI clients. It connects to model providers including OpenRouter, OpenAI, Anthropic, DeepSeek, Gemini, OpenAI-compatible gateways, Ollama, and vLLM using your own API keys. You can switch models mid-session without losing conversation history.

Do I need API keys to use DeepCode?+

Yes, you need an API key from at least one supported model provider (OpenRouter, OpenAI, Anthropic, DeepSeek, etc.) to use DeepCode. Configure your provider and key via 'deepcode provider set' in CLI or Settings → AI providers in Desktop/Web before starting your first session.

Is DeepCode free to use?+

DeepCode itself is free and open-source under the MIT License. However, you pay for the model provider API usage (OpenAI, Anthropic, etc.) according to each provider's pricing. You can use free local models via Ollama or vLLM to avoid cloud costs.

What makes DeepCode different from other coding agents?+

DeepCode provides durable sessions that survive restarts and interface switches, Goal-driven Loop Engineering that continues work across multiple turns, evidence-driven verification with actual test execution and results, and specialized Paper2Code workflows for research reproduction. It also offers parallel agent execution, reusable Skills, MCP integration, and scheduled Automations for repeatable engineering work.

How do I install DeepCode?+

Open the source repository on GitHub and follow its README. DeepCode is a agent — MCP Agents Market links you directly to the official repo.

Is DeepCode free?+

DeepCode is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches

Developer Tools AgentsCode Review AgentsResearch AgentsBest MCP ServersBest SkillsBest PluginsTop rated listingsClaude Code toolsBrowse all categories