Does DeepSeek Harness work with Standard Compute?

Yes. DeepSeek Harness supports a custom OpenAI-compatible base URL, so it works with Standard Compute. Set the base URL to https://api.stdcmpt.com/v1, paste your API key, and set the model to "standardcompute".

What base URL do I use for DeepSeek Harness?

Use https://api.stdcmpt.com/v1 as the OpenAI-compatible base URL in DeepSeek Harness, with your Standard Compute API key and the model "standardcompute".

Do I need to change my code or model names?

No code changes. The model "standardcompute" routes each request to a frontier or an efficient model, so you don't pick or update model names; an exact model ID from the Models page pins one model instead.

← All integrations

OpenAI-compatible · Custom base URL · Agent harness

Use Standard Compute with DeepSeek Harness

DeepSeek's open-source agent harness (dsh) for coding, research and everyday work, run as a desktop app or a local web UI, with custom model APIs in three protocols. Paste three settings into dsh: the base URL, your Standard Compute API key and the model.

dsh's Custom model API speaks OpenAI Chat Completions, OpenAI Responses or Anthropic Messages, one protocol per provider. The Models page saves to ~/.dsh/profiles/web/cordis.patch.yml, and changes apply on the next request without a restart.

Connect dsh

Eligible new accounts can try $0.25 of compute, with no card. See trial terms →

Your connection settings

Copy these values, then put them where dsh expects them (steps below).

Base URL

https://api.stdcmpt.com/v1

API Key

Your Standard Compute key

Model

standardcompute

Where to set them in dsh

Open Settings → Models, choose Add model provider and switch the card to Custom model API.
Enter a Provider ID such as standardcompute (it is permanent), the base URL https://api.stdcmpt.com/v1, API protocol OpenAI Chat Completions and your Standard Compute API key.
Add the model standardcompute, or choose Fetch available models and tick exact IDs such as anthropic/claude-sonnet-5.5, then save and pick it in the session's model picker.

dsh docs ↗

What changes in dsh

Billing: dsh requests draw on one monthly compute budget, with plans from $19 a month. There is no per-token bill. When the budget is used, requests stop until renewal or an upgrade, unless you turn on pay as you go.
Models: The model standardcompute picks a frontier or an efficient model for each request. An exact model ID from the Models page pins one model instead.
Good to know: DeepSeek Harness is free and MIT-licensed, in developer preview with compatibility-breaking changes expected. Its built-in route is a DeepSeek account or API key; a custom provider bills through that provider instead.