Execution model
Brainlet is designed for local or company-controlled execution, so project indexing can happen in the environment where the repository already lives.
Security
Project intelligence should not require handing over the whole project.
Brainlet is designed around a local-first security posture: compute project intelligence where the code already lives, then let teams decide which model receives which task context.
Repository handling
The website does not claim that every possible integration is offline. The security boundary depends on how a team connects Brainlet to its chosen LLM and workflow.
LLM choice
Teams can connect local models, private hosted models, or frontier APIs according to their own policy. Brainlet's role is to prepare project intelligence before the model acts.
Data minimization
The product direction is to send the model only the project facts needed for a task, rather than uploading an entire repository as raw context.
Controls
The practical security model.
- 01 Run indexing in a local or company-controlled environment.
- 02 Choose which LLM endpoint receives task context.
- 03 Use project intelligence instead of broad raw-file prompt stuffing.
- 04 Review generated findings before code changes are accepted.