#99th of 83 in Agent and MCP starters
soduwork
Team workspace that routes requests to agent harnesses with AI review
- Stars
- 6.2k
- License
- MIT
- Last commit
- Nov 2026
- Last release
- Nov 2026
- Language
- JavaScript
3,945 on 27 Oct 20266,213 on 17 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
SoduWork is a Hub, admin console and desktop client that dispatches requests to agent harnesses such as Pi, OpenCode, Qwen Code, Kimi Code and Codex. Each request carries acceptance criteria, a separate AI review checks them, and a person signs off. Working methods can be saved as versioned, reviewable plugins shared across a team. The README describes it as a public beta.
Who it is for: Teams that want a shared workspace to assign, run and review coding-agent work
Strengths
- Mixes harnesses and models per step, with a different model reviewing the work
- Plugins are versioned, reviewed before sharing, and can be granted or revoked
- Roles, usage and cost tracking, and an audit log are built in
- Docker deploy script starts Hub, admin console and PostgreSQL on one server
Weaknesses
- Public beta; README advises evaluating before production use
- Running from source needs Python 3.12+, Node.js 22.19+ and Git
- Hub usage accounting and update polling cannot be disabled
- Runners are not a general sandbox; isolation depends on harness and environment
What it needs
- JavaScript, OpenAI-compatible chat models, DeepSeek, Codex (own OpenAI account), self-hosted models
- Needs Python 3.12+, Node.js 22.19+, Git, PostgreSQL (Docker deploy)
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Agent and MCP starters
See all 83| Rank | Project | Score |
|---|---|---|
| 1 | deepseek-harnessPlugin-based agent harness from DeepSeek with a local Web UI | 98 out of 100 |
| 2 | composioTypeScript and Python SDKs that give agents authenticated tools across 1000+ apps | 93 out of 100 |
| 3 | conductorDurable workflow engine that orchestrates microservices and LLM agents from JSON | 91 out of 100 |
| #4 | a2uiProtocol and renderers for agents to stream native UI instead of text | 88 out of 100 |
| #5 | mcp-useTypeScript framework for building MCP servers with React Views and an Inspector | 88 out of 100 |