Skip to main content
DeepSeek Harness is an open-source agent harness with a local Web UI. After starting the local server, you can configure A2Agent from Settings → Models, choose a model, select a workspace, and run tasks directly in your browser.

​
Prerequisites

DeepSeek Harness is currently in the Developer Preview stage. Future releases may include breaking changes.

​
Setup

1

Start the local Web UI

Run the following command:
The --no-open option prevents DeepSeek Harness from trying to open a browser automatically.When the server starts, it normally listens at:
Keep the terminal running while you use the Web UI.
2

Open DeepSeek Harness in your browser

Open:
If localhost is not available, try:
3

Open the model settings

In the DeepSeek Harness Web UI, go to:
Choose:
This option is used to connect an OpenAI-compatible gateway or another provider that is not included in the built-in catalog.
4

Configure the A2Agent provider

Enter the following values:Use this Base URL exactly:https://api.a2agent.me/v1
5

Add A2Agent models

Under Model catalog, choose:Fetch available modelsDeepSeek Harness will query the OpenAI-compatible models endpoint and display the models available through A2Agent.Select the models you want to use, then save the provider.If model discovery does not work, add model IDs manually. Use the exact IDs shown on the A2Agent Models page.
6

Return to the homepage and select a model

After saving the A2Agent provider, return to the DeepSeek Harness homepage.Open the model selector and select one of the available models.The selected model becomes the model used for new sessions.

​
Stop and restart the Web UI

To stop the local server, return to the terminal running DeepSeek Harness and press:
To start it again, run:
Then open:

​
Troubleshooting

DeepSeek Harness reached A2Agent, but the API key was rejected.
  • Confirm that the key is a valid A2Agent API key.
  • Make sure the complete key was entered.
  • Check for accidental spaces or line breaks.
  • Do not use an API key from another provider.
  • Reopen Settings → Models and update the saved credential.
Confirm that the Base URL is exactly:https://api.a2agent.me/v1A2Agent uses the OpenAI-compatible protocol. Select:OpenAI CompletionsDo not use an Anthropic or provider-specific protocol for this endpoint.
Check that:
  • the provider was saved successfully;
  • the Provider ID is a2agent;
  • the Base URL is correct;
  • the API key is valid;
  • at least one model was added; and
  • the model ID matches the current A2Agent model list.
Confirm that the process is still running and that port 3080 is not already in use.Try:
If the terminal reports a JavaScript heap out-of-memory error while npx is starting the application, close other Node.js processes and retry with a larger Node.js heap:
DeepSeek Harness is a local Web UI. It must remain running in the terminal while you access it from your browser.