springbrand deepseek harness
DeepSeek Harness (dsh) is an open-source agent harness from DeepSeek AI, released under the MIT license and currently in developer preview (v0.1.0-rc.6). It is built on the Cordis plugin kernel, where every capability—models, tools, sessions, sandboxes, and the agent loop itself—is implemented as a replaceable plugin. The framework is written in TypeScript and runs on Node.js 22.19+ or 24+, starting a local web UI with a single command: npx @deepseek-ai/dsh web. Unlike a model, DeepSeek Harness does not train, host, or replace models. It provides the layer around a model that reads the workspace, calls tools, enforces approval policies, maintains an append-only session log, and keeps the run moving. The model is configured through any provider, while Cordis handles mounting, unmounting, and dependency tracking of plugins, so registrations unwind when a plugin is unloaded. Users extend dsh by adding plugins rather than changing a core; a swap can redirect filesystem and subprocess capabilities to a remote sandbox. It ships a working web UI and a headless runner, but its primary surface is the plugin tree, where profiles, bundles, and individual configuration rows can be selected. As a developer preview, compatibility-breaking changes are expected, and the CLI rejects serving the web UI over a network to non-local hosts.
- Category
- AI Agent
- Pricing
- Unknown
Problem solved
DeepSeek Harness solves the need for an open, inspectable agent runtime by giving a model a workspace, tool registry, sandbox and approval policy, an append-only session log, and a persistent loop. It lets teams own and extend their agent stack through swappable plugins instead of relying on a closed hosted service, enabling customization of model adapters, sandboxes, storage backends, and tools by configuration.
Use cases
- Building custom agent harnesses with swappable model, tool, and sandbox plugins
- Running DeepSeek Harness Web UI locally via npx @deepseek-ai/dsh web
- Developing and publishing dsh plugins under the dsh-plugin topic
- Evaluating models with a stable, minimal tool surface across runs
- Embedding an inspectable agent loop in internal platforms
- Swapping storage backends or sandboxes by configuration
Audience
- Developers
- ML Engineers
- Data Scientists
- Researchers
Pros
- MIT-licensed and open source with no waitlist or beta invite
- Plugin architecture built on Cordis with swappable models, tools, sandboxes, and storage
- Runs locally on your own machine with a one-command start
- Provides both a web UI and a headless runner
- Extensible via community plugins under the dsh-plugin topic
Cons
- Developer preview with compatibility-breaking changes expected
- Not suitable for production deployments requiring API stability
- Web UI cannot be served to a network (CLI rejects --host 0.0.0.0)
- Requires Node.js 22.19+ or 24+ and a model API key
- Not a hosted or managed service; users must run and maintain it themselves
Official website: https://springbrand.ai/deepseek-harness
Canonical LambdaBase page: https://www.lambdabase.com/tools/springbrand-deepseek-harness