NativeAIHub

Agent Mode

All plans (Free: 1,000 requests/day, Standard: 1,500/day, Enterprise: 2,000/day)3 min read
1

Describe the task

Write a natural language description of what you want done. Be specific about the scope and constraints.

2

Agent gathers context

The agent reads relevant files, analyzes project structure, and builds understanding of the codebase.

3

Review the Outline

The agent generates a high level Outline showing the approach, files to modify, and key decisions. Approve or refine before coding starts.

4

Execute with Finish Changes

The agent writes code across files, then Finish Changes auto completes remaining edits to imports, tests, and related code throughout the project.

5

Review inline diffs

Every change appears as an inline diff in your editor. Approve, reject, or provide feedback for iteration.

6

Iterate until complete

The agent incorporates your feedback and continues until the task is done and all tests pass.

0/day

Free tier agent requests

0/day

Standard tier agent requests

0/day

Enterprise tier agent requests

📋
OutlinesThe agent generates a high level plan showing the approach, files to modify, and key decisions before writing any code. Review and refine the plan to ensure alignment.
🔄
Finish ChangesAfter initial edits, auto completes remaining related changes across all affected files: imports, tests, configs, and call sites throughout the project.
👤
Human in the LoopThe agent pauses after each significant change to show diffs and wait for your approval. Full control over every change that enters your codebase.
âš¡
Auto ApproveThe agent applies changes automatically without pausing. Maximum speed for trusted, well scoped tasks. Review all changes after completion.

Getting the best results from agent mode

Start with human in the loop mode until you are comfortable with the agent's quality on your codebase. Be specific in task descriptions: "Add form validation to the signup page: email format check, password minimum 8 characters, and display inline error messages" produces better results than "improve the signup form." For large refactors, break the work into smaller, focused agent tasks rather than one massive request.