Agent is the default and most autonomous mode in Teta, designed to handle complex coding tasks with minimal guidance. It has all tools enabled to autonomously explore your codebase, read documentation, browse the web, edit files, and run terminal commands to complete tasks efficiently.

​
Agent’s Capabilities

Autonomous Operation

Independently explores your codebase, identifies relevant files, and makes necessary changes

Full Tool Access

Uses all available tools to search, edit, create files, and run terminal commands

Contextual Understanding

Builds a comprehensive understanding of your project structure and dependencies

Multi-step Planning

Breaks complex tasks into manageable steps and executes them in sequence

​
Agent’s Workflow

Much like a human, Agent mode follows a systematic approach to completing tasks:

1

Understand Request

Agent mode analyzes your request, and the context of the codebase to fully comprehend the task requirements and goals.

2

Explore Codebase

The agent may search through your codebase, documentation, and the web to identify relevant files and understand the current implementation.

3

Plan Changes

Based on the analysis, the agent breaks down the task into smaller steps and plans the changes, learning from the available context as it goes.

4

Execute Changes

The agent makes the necessary code modifications according to the plan across your entire codebase, also potentially suggesting new libraries, terminal commands to run or steps you should make outside of Teta.

5

Verify Results

The agent will confirm the changes look correct after being made by applying them. If it notices any issues or linter errors (when supported by the language), it will attempt to fix them.

6

Task Complete

Once the agent is happy, it yields back and summarizes the changes it made.

Agent mode creates stores any changes in internal commits, allowing you to revert if needed.
Learn more about commits.