FAQ

This page answers common questions about Skynet. Each answer links to the page with the full details.

General

What is Skynet?

Skynet is a platform that optimizes prompts, programs, agents and workflows against your own data and metric. You upload a dataset, describe the task, and start a run. The run tries new versions and scores them on held-out data. To start, read the Quickstart.

What can Skynet optimize?

Skynet is Optimize Anything. It optimizes prompts, code, text and DSPy programs in one product. The artifact kind decides which engines can work on it:

• Prompt, Code and Text artifacts use the native engines: GEPA, Best-of-N, Meta-Harness, AutoResearch, AutoSaddler and ShinkaEvolve.
• DSPy programs use the DSPy optimizers: DSPy GEPA, DSPy MIPROv2 and DSPy BootstrapFewShot.

Auto combines the engines that fit your task. See How optimization works and Engines.

Which languages does the interface support?

The interface has 24 languages. Hebrew is the base language, and the layout supports right-to-left text. Hebrew docs are at /docs/he.

Is Skynet open source?

Yes. The code uses the AGPL-3.0 license. If you offer a changed version of Skynet as a service, the license requires you to share your changes.

Data

Which file types can I upload?

You can upload CSV, JSON, .xlsx and .xls files. For Excel files, Skynet reads only the first sheet. See Prepare a dataset.

My data has no labels. What can I do?

Use the tagger. You can label rows yourself, or let an AI model label them while you review its work. See Label data with the tagger.

Who can see my runs and datasets?

Set Visibility to Private when you submit a run to keep it out of public search. New datasets are Restricted: only you and the people you invite can open them. See Sharing and Security.

Do model providers train on my prompts?

By default, no. The default Model data privacy option is No training. It sends prompts only to providers that do not train on them and do not keep them. This setting applies to calls that Skynet bills to your balance. Calls on your own key follow your OpenRouter account's privacy settings. See Security.

Billing

How does Skynet charge for runs?

Skynet uses a prepaid balance. Hosted model usage, Skynet fees and sandbox usage are metered. Before paid work starts, Skynet reserves headroom in your wallet and in the run budget. See Billing.

Can I use my own model provider key?

Yes. With your own key (BYOK), the provider charges your key directly for model calls. Your Skynet balance still pays the platform fee and sandbox usage. Skynet encrypts your key and never shows it again. See Billing and Security.

Can I run Skynet without paying for model calls?

The Mac app runs Skynet on your own computer. It sends model calls through Claude Code or Codex, signed in on your Mac. Your CLI plan pays for those calls, so they add $0 to the run.

Tools and integrations

Can I control Skynet from my own AI client?

Yes. Skynet has an MCP server at https://mcp.skynetml.com/mcp. You sign in with OAuth. The server has these tools: check_run, start_run, get_run, list_runs and stop_run. See MCP.

Is there an API?

Yes. Generate an API token in Settings, then send it as Authorization: Bearer <token>. Each user has one valid token at a time. See API.

Can the agent panel delete my runs?

Only if you allow it. In the default Ask mode, the agent asks before every action. In Automatic mode, it still asks before destructive actions, such as delete or submit. See Agent panel.