Choose your AI model
Redline Desk runs on your Windows 10 or 11 machine. You pick the model that fits your privacy needs and report complexity. You choose which model receives your prompts.
Three ways to run intelligence
Cloud AI provider
Connect a standard API endpoint to use a large language model hosted elsewhere. Your prompts travel to that provider, so check your client contract for data residency rules. This option works well when you need a larger hosted model to suit long drafts or difficult severity alignment. You control the endpoint URL and keep your API keys on your PC. Secrets in tool output are redacted before they leave your machine.
Local PC model
Run a model directly on your Windows computer using available hardware resources. Nothing leaves your machine when you use this mode, which is ideal for highly confidential pentest reports. Your working folder and chats stay local. You can check the endpoint with the Test button, which shows response time and the context window.
Built-in model
Use the default model that comes with the application for immediate use without extra configuration. This option is convenient for quick tasks like formatting a single finding or checking a diff. It is ready to use without your own endpoint. You can switch to a cloud or local model later if the report demands higher complexity.
Verify your connection
Press the Test button to check your chosen endpoint. The tool shows the response time and confirms the context window size. You see the result immediately before starting any writing tasks.
Match model power to the task
Send hard steps to a bigger model using run_on_model for complex logic. Use smaller models for small edits or simple file writes. The task orchestration can parallelize sub-agents for drafting multiple findings at once. You decide which model handles which part of the report.
Quality varies between models, so always review the diff before accepting. Use Ask first mode to pause before every action that changes something. You see the exact text added to your report. This ensures every finding matches your notes and evidence exactly.
