Skip to content

Configuration Guide ​

Set up NimoteCode around the work you actually do. A small, tested configuration is safer and more useful than a large catalog of profiles you do not recognize under pressure.

AI: keep a primary and a backup profile ​

Add the provider endpoint, credential and model you plan to use; then verify it in AI Chat before relying on Agent. A primary profile and a tested fallback help balance reliability and cost. Choose lower randomness for precise code tasks, use streaming when it improves responsiveness, and set context behavior deliberately.

AI model configuration on NimoteCode for phone — phone
AI model configuration on NimoteCode for phone
AI edit configuration on NimoteCode for phone — phone
AI edit configuration on NimoteCode for phone

SSH: name the environment, then set the boundary ​

Create one profile per environment, with a recognizable name, host, port, user and authentication method. Confirm the workspace root after connecting. Prefer key-based access and least-privilege accounts for sensitive hosts; never assume a familiar profile still points to a safe environment.

Tasks: record known procedures, not guesses ​

Create clear tasks for build/test, format/lint, health checks and log inspection. Group them by purpose and make any restart, deployment or data-changing command unmistakable. Tasks run through an SSH workflow, so validate access and current directory before execution.

Ready-to-work checklist ​

  1. AI Chat returns a response from the intended profile.
  2. The SSH profile reaches the intended host and project root.
  3. One low-risk terminal command works in the workspace.
  4. One well-understood task runs successfully.
  5. If you use Pro Git workflows, Source Control can read the repository state before you write anything.

Continue with Settings · AI Chat and Agent · Security and Safety