Skip to content
Get packs

Guide · Open source

How to use dsh-our-free-model: install and run free models in dsh

Updated · written with AI from the project's own documentation and checked against our catalogue data

dsh-our-free-model is an MIT-licensed JavaScript plugin for DeepSeek Harness (dsh), an AI agent host. Once installed in dsh, it adds hosted models such as DeepSeek V4.1 Flash and Kimi K3 to the model picker without a login, sign-up or API key. This guide shows how to install it from a local source directory, what to expect from a beta plugin, and when to choose a different tool.

Our Free ModelEbony-Vinyl/dsh-our-free-model · ★ 56.6k · MIT

Quick start

  1. 1Confirm your dsh version. Make sure your DeepSeek Harness version is within 0.1.5 to 0.1.7-rc.2, as listed in the README.
  2. 2Get the plugin source. Download the Release package or clone the repository. Do not use the npm package with the same name.
  3. 3Install into the web profile. Run dsh plugin --profile web add with the absolute path to the plugin folder, replacing the profile name if needed.
  4. 4Restart dsh. Restart the app once so the plugin loads.
  5. 5Pick a free model. Open the model picker and select a model under the Our Free Model group, such as DeepSeek V4.1 Flash or Kimi K3.
  6. 6Test the local port (optional). Open the local forwarding section in the settings page and run the curl example to confirm other tools can reach the models.

Check that you can use this plugin

The plugin only works inside DeepSeek Harness (dsh). It is not a standalone tool and does not run in other hosts.

The README lists the dsh kernel versions 0.1.5 to 0.1.7-rc.2 as supported. Check your dsh version against that range before you install.

  • You already run DeepSeek Harness (dsh).
  • Your dsh version falls within 0.1.5 to 0.1.7-rc.2.
  • You can open a terminal and access the dsh command.

Install the plugin from a local source directory

Download or clone the repository, then point dsh at the folder. Replace the path with the absolute path to your copy. The command adds the plugin to the web profile.

After the command finishes, restart the app once. Replace the profile name if you use a different one.

  • Use the Release package from the repository or a local source directory.
  • Do not install it from npm. The README warns that the npm package named dsh-our-free-model is third-party and differs from this repository.
dsh plugin --profile web add /绝对路径/dsh-our-free-model

Choose a model in dsh

After restart, open the model picker in dsh. Models from the free lane appear under the Our Free Model group. Models the gateway reports as region-limited are listed separately under a region-limited group.

The README says the model list refreshes from upstream, and availability depends on your own network exit. A model that is missing one day may return later.

  • Free lane: models usable from your current network exit.
  • Region-limited: models the upstream does not allow in your region, kept visible but separated.
  • Settings page: the plugin's settings screen shows model availability, vision support, context window and output limits.

Use the local OpenAI-compatible port

The plugin includes a local OpenAI-compatible port, so other tools on the same machine can call the same models. Tools connect with a base URL and key. The settings page has a local forwarding section that shows the base URL and key controls, plus a curl example you can run directly.

The README also states that the local gateway is for your own machine. Do not expose it to other people without understanding the risk.

  • Find the base URL and API key in the plugin's local forwarding section of the settings page.
  • Use the curl example there to test the connection before configuring other tools.
  • Only use this port for prompts you are comfortable sending to the free upstream pools.

Know the privacy and stability limits

The README says the Kilo free pool may record prompts to improve the service. Do not send sensitive or confidential code or text through the free models.

The project is marked beta. Model lists refresh from upstream, so models and their availability can change without notice. Use it for testing and comparison, not for work that needs guaranteed uptime.

  • Never send secrets, customer data or confidential source code.
  • Expect model names and availability to change.
  • Keep a second tool ready if you depend on the output.

What people use it for

  • Trying DeepSeek V4.1 Flash or Kimi K3 inside dsh without creating accounts.
  • Comparing several hosted models from one agent tool.
  • Pointing a local tool at an OpenAI-compatible base URL on your own machine.
  • Testing agent workflows where occasional outages or model changes are acceptable.

Frequently asked questions

Is dsh-our-free-model free to use?

The README says the models are free and have no usage cap. The free models come from the OpenCode Zen and Kilo AI public gateways, and the upstream providers set the terms.

Do I need an API key or account?

No. The README says you only install the plugin in dsh, with no login, sign-up or API key required.

How do I install dsh-our-free-model?

Run dsh plugin --profile web add with the absolute path to a local copy of the repository, then restart dsh. The README says not to install the npm package with the same name, since it is third-party.

Why does the model list change?

The README says the list refreshes from upstream and availability is tested from your own network exit. Models can disappear or return as upstream routing changes.

Is it safe to send private code through the free models?

The README advises against it. The Kilo free pool may record prompts to improve the service, so keep sensitive or confidential content out of these models.

How does it compare with Ollama, LiteLLM or Open WebUI?

Ollama runs models locally, LiteLLM is a gateway for many providers, and Open WebUI is a browser interface. This plugin only works inside dsh and gives access to hosted free models. Choose it for quick trials inside dsh, and choose the others if you need local models, provider routing or a full chat interface.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use dsh-our-free-model if you already run DeepSeek Harness and want to try hosted models without accounts or keys, and you can accept a beta plugin with changing availability. Pick another option if you handle sensitive data, need guaranteed uptime, or do not use dsh.

Install options & downloads