Analysis updated 2026-08-15
Run a local web UI to interact with the DeepSeek agent framework.
Build and share custom plugins that extend the agent's capabilities.
Clone and contribute to an open-source agent framework during its early preview phase.
| deepseek-ai/deepseek-harness | immich-app/immich | angular/angular | |
|---|---|---|---|
| Stars | 99,325 | 99,828 | 100,060 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | easy | hard | moderate |
| Complexity | 3/5 | 4/5 | 4/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Simplest path is a single npx command that downloads and launches a web UI, cloning from source requires pnpm.
DeepSeek Harness, or dsh, is an open-source agent framework created by DeepSeek AI. Its defining design principle is that "everything is a plugin," meaning the system is built so that its capabilities can be extended or swapped through modular add-ons rather than being locked into a fixed set of features. The project is built on top of a system called Cordis, which the README links to an academic paper on composability. The project is written in TypeScript and is currently labeled as a "developer preview." The README is clear that the codebase is iterating rapidly and warns that there will be compatibility-breaking changes. This means anyone building on top of it right now should expect that updates might require adjustments to their code or plugins down the line. Getting started involves running the tool through Node.js. The simplest path is using the npx @deepseek-ai/dsh web command, which downloads and starts a web-based user interface on your local machine at a specific port. Alternatively, developers can clone the repository directly, install dependencies using a tool called pnpm, build the project, and run it from source. The README points to separate documentation files for using the web interface and understanding the architecture. The project encourages community involvement through GitHub Discussions and a Discord server. Developers who create plugins are encouraged to tag their repositories with the dsh-plugin topic so others can find them. There is a separate contributing guide, a development guide, and a specific file with instructions for automated coding agents. The project uses the MIT license and includes a file disclosing the licenses of third-party dependencies.
An open-source agent framework from DeepSeek AI where every feature is a modular plugin you can swap or extend, built in TypeScript and currently in an early, fast-changing developer preview.
Mainly TypeScript. The stack also includes TypeScript, Node.js, pnpm.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.