Pricing
Two ways to run Darkshift
Self-serve gets you a working factory today: $99 per user per month, with model usage billed as you go. If you need it inside your own infrastructure, cloud or on premises, a custom implementation is scoped and quoted individually.
Self-serve
$99 per user, per month
The platform fee for each user on your team.
- Connect your GitHub repos and run tasks with Codex, Claude Code, Grok Build, or Kimi
- Bring your own API keys or Codex subscription, or pay provider token rates with no markup
- Run your app on Darkshift cloud
- Secure cloud sandboxes for agents
- Evidence-backed PRs that deliver
- Connect Linear to feed work to Darkshift
- Team collaboration
Custom implementation
Custom scoped and quoted individually
Darkshift deployed on your own infrastructure and adapted to your setup.
- Runs in your cloud, or fully on premises
- Customized harnesses, integrations, and credential mapping
- Your models and your keys if you want them
Model usage
How model usage is billed
The $99 covers the platform, not the AI. Model usage is paid one of three ways.
01
Metered, the default
Nothing to configure. Usage is billed at each provider's published token rates and itemized on your monthly invoice. No markup.
02
Your API keys
Add your Anthropic, OpenAI, Azure OpenAI, or xAI key. The provider bills you directly; Darkshift charges nothing for models.
03
Your Codex subscription
Connect a ChatGPT plan with Codex and Codex tasks run on it, covered by what you already pay OpenAI.
Claude Code subscriptions can't be attached to outside products, so Anthropic models run on your API key or the metered default. Kimi has no key option yet and runs metered only.
FAQs
Common questions
What does the self-serve plan include?
The $99 per user per month covers platform access: tasks against your connected GitHub repos, isolated sandboxes, recorded browser QA, and pull requests as the output. It is not an AI allowance; model usage is billed separately, one of three ways. You can cancel anytime.
How do we pay for model usage?
Three ways. Default: Darkshift meters your usage and bills it at each provider's published token rates, itemized, with no markup. Or add your own API key (Anthropic, OpenAI, Azure OpenAI, or xAI) and the provider bills you directly. Or connect a ChatGPT plan with Codex and Codex tasks run on that subscription. Claude Code subscriptions can't be attached to outside products, so Anthropic models use an API key or the metered default, and Kimi runs metered only for now.
Can we run Darkshift on premises?
Yes, as a custom implementation. The factory is deployed inside your infrastructure, in your cloud or fully on premises, and adapted to your workflow: your integrations, your credential mapping, and your models if you want them.
Why is there no price for the custom implementation?
Because there is no standard custom deployment. The factory runs on your infrastructure with your integrations, so scope and price are quoted individually before you commit to anything. The standard deployment is the self-serve plan, priced above.
Does our code or our credentials ever leave our infrastructure?
On a custom implementation the factory runs in your cloud or on premises, and credential brokering keeps secrets out of the sandboxes where possible: agents do their work without holding a real key. On the self-serve plan, tasks run in isolated sandboxes on Darkshift cloud.
See it run on your repo
Book 30 minutes and watch Darkshift clear a real ticket from your repo, live. Book a demo.