Creating Tasks

Create coding tasks in Terragon

Starting a New Task

Click on Home from the sidebar

Select your repository, your branch and model near the prompt box

Write in your prompt and include any context

Submit

Adding Context

Reference specific files with @:

Type @ to trigger autocomplete

Start typing the filename

Select from suggestions

Refactor the authentication logic in @src/auth/login.ts to match the patterns used in
@src/auth/register.ts

Attach images, PDFs, and text files:

Click the icon or drag-and-drop

Paste images from clipboard with CMD + V / Ctrl + V

Supports images (PNG, JPG, GIF, WebP, etc.), PDF, and text files (TXT, JSON, etc.) up to 10MB each

Dictate complex requirements:

Click the icon

Prompt at the speed of thought

Review and edit the transcription

Include links for additional context:

  Fix the bug described in this issue:
  https://github.com/myorg/myapp/issues/234

Best Practices

Write Descriptive Prompts: The more context you give the AI agent, the better. Meaty task descriptions often require little or no follow-up.

Think Parallel: Terragon shines when you split your work between multiple async AI agents—whether it's fixing bugs, starting new features, or doing background research.

Be Specific

❌ Fix the bug

✅ Fix the authentication timeout bug where users are logged out after 30 minutes of inactivity. The issue is likely in @src/auth/session.ts where we handle token refresh.

Provide Context

❌ Add tests

✅ Add unit tests for the payment processing module in @src/payments/. Cover successful transactions, failed payments, refund processing, and webhook handling. Use existing patterns from @src/payments/stripe.test.ts

Set Clear Goals

❌ Improve performance

✅ Optimize dashboard loading time by implementing lazy loading for charts, adding pagination to the activity feed (10 items per page), and using React.memo for expensive components. Target: under 1 second load time.

Multi-Agent Mode

You can run multiple AI agents from a single prompt to compare different approaches or evaluate which model is best suited for your task:

Each selected agent will create a separate task, working independently on your prompt. This is particularly useful when:

You're exploring different implementation approaches
You're unsure which model will perform best for your specific task
You want to compare outputs from different AI agents
You're prototyping and want to see multiple solutions quickly

After Submission

Your task will:

Provision a development environment

Clone the repository, install dependencies and run your setup script

Create acceptance criteria and a definition of done when automated acceptance is enabled

Execute the requested changes

Run tests if specified

Create a pull request when complete

Wait for PR checks, try to fix failures, and run final review/testing when automated acceptance is enabled

Learn more in Automated Acceptance.

Creating Tasks

Create coding tasks in Terragon

Starting a New Task

Click on Home from the sidebar

Select your repository, your branch and model near the prompt box

Write in your prompt and include any context

Submit

Adding Context

Reference specific files with @:

Type @ to trigger autocomplete

Start typing the filename

Select from suggestions

Refactor the authentication logic in @src/auth/login.ts to match the patterns used in
@src/auth/register.ts

Attach images, PDFs, and text files:

Click the icon or drag-and-drop

Paste images from clipboard with CMD + V / Ctrl + V

Supports images (PNG, JPG, GIF, WebP, etc.), PDF, and text files (TXT, JSON, etc.) up to 10MB each

Dictate complex requirements:

Click the icon

Prompt at the speed of thought

Review and edit the transcription

Include links for additional context:

  Fix the bug described in this issue:
  https://github.com/myorg/myapp/issues/234

Best Practices

Write Descriptive Prompts: The more context you give the AI agent, the better. Meaty task descriptions often require little or no follow-up.

Think Parallel: Terragon shines when you split your work between multiple async AI agents—whether it's fixing bugs, starting new features, or doing background research.

Be Specific

❌ Fix the bug

✅ Fix the authentication timeout bug where users are logged out after 30 minutes of inactivity. The issue is likely in @src/auth/session.ts where we handle token refresh.

Provide Context

❌ Add tests

Set Clear Goals

❌ Improve performance

Multi-Agent Mode

You can run multiple AI agents from a single prompt to compare different approaches or evaluate which model is best suited for your task:

Each selected agent will create a separate task, working independently on your prompt. This is particularly useful when:

You're exploring different implementation approaches
You're unsure which model will perform best for your specific task
You want to compare outputs from different AI agents
You're prototyping and want to see multiple solutions quickly

After Submission

Your task will:

Provision a development environment

Clone the repository, install dependencies and run your setup script

Create acceptance criteria and a definition of done when automated acceptance is enabled

Execute the requested changes

Run tests if specified

Create a pull request when complete

Wait for PR checks, try to fix failures, and run final review/testing when automated acceptance is enabled

Learn more in Automated Acceptance.

Creating Tasks

Starting a New Task

Adding Context

Multi-Agent Mode

After Submission

On this page

Creating Tasks

Starting a New Task

Adding Context

Multi-Agent Mode

After Submission

On this page