Prompt Studio: Complete Guide & Tutorial

Introduction to Prompt Studio

In the rapidly evolving landscape of software development, AI coding agents have emerged as powerful assistants capable of writing, reviewing, and refactoring code. However, managing these agents at scale—especially when running multiple tasks in parallel—presents a significant challenge. Enter Prompt Studio, an open-source command-line interface (CLI) and dashboard designed to help developers delegate tasks to AI coding agents while maintaining full control over the workflow.

Prompt Studio allows you to orchestrate parallel agents, scale your agentic coding workflows, and keep a watchful eye on every task your agents perform. Whether you are a solo developer working on a complex refactoring project or a team lead managing multiple feature implementations, Prompt Studio provides the infrastructure to coordinate AI agents efficiently. It supports macOS, Linux, and Windows, making it accessible to virtually any development environment.

This tutorial will guide you through everything you need to know to get started with Prompt Studio, from installation to advanced workflow orchestration. By the end, you will be able to set up your first parallel agent tasks and scale your coding productivity without losing visibility or control.

Getting Started with Prompt Studio

System Requirements and Installation

Before you begin, ensure your system meets the following requirements:

• Operating System: macOS 10.15+, Ubuntu 20.04+ (or any modern Linux distribution), or Windows 10/11 with WSL2 or native support.
• Node.js: Version 18 or higher is required.
• Git: Version 2.30 or higher for repository operations.
• API Key: An active API key for your preferred AI model provider (e.g., OpenAI, Anthropic, or a local model via Ollama).

To install Prompt Studio, open your terminal and run the following command:

For macOS and Linux:
curl -fsSL https://prompt.studio/install.sh | sh

For Windows (using PowerShell):
iwr -useb https://prompt.studio/install.ps1 | iex

Alternatively, you can install via npm if you prefer:

npm install -g @prompt-studio/cli

After installation, verify that Prompt Studio is correctly installed by typing:

prompt-studio --version

You should see the current version number displayed in your terminal.

Initial Configuration

Once installed, you need to configure Prompt Studio with your AI provider credentials. Run the setup command:

prompt-studio init

The CLI will guide you through a series of prompts:

• Select your AI provider: Choose from OpenAI, Anthropic, Google Gemini, or Ollama (for local models).
• Enter your API key: Paste your API key when prompted. The key is stored locally in a configuration file at ~/.prompt-studio/config.json.
• Set default model: Specify which model to use by default (e.g., gpt-4, claude-3-opus, or codellama).
• Define working directory: Choose where your project files reside. This is typically your current project folder.

After configuration, you can test your setup with a simple command:

prompt-studio run "Explain what this README does"

If everything is working, you will see a response from the AI agent in your terminal.

Key Features of Prompt Studio

Open-Source CLI and Dashboard

Prompt Studio is fully open-source, meaning you can inspect, modify, and contribute to the codebase. The CLI provides a powerful terminal-based interface for power users, while the web dashboard offers a visual representation of your agent workflows. The dashboard runs locally on http://localhost:3000 and can be launched with the command prompt-studio dashboard.

Parallel Agent Orchestration

One of the standout features is the ability to run multiple AI agents simultaneously. Instead of waiting for one agent to finish before starting the next, you can define a batch of tasks and let Prompt Studio execute them in parallel. This is particularly useful for:

• Refactoring multiple files at once.
• Generating unit tests for several functions simultaneously.
• Reviewing pull requests in parallel.
• Running code generation for different modules concurrently.

Scalable Workflow Management

Prompt Studio is built to scale. You can start with a handful of tasks and expand to hundreds of agents working on a single codebase. The tool handles rate limiting, retry logic, and error handling automatically, so you do not have to worry about API throttling or failed requests disrupting your workflow.

Granular Control and Monitoring

Maintaining control over AI agents is critical. Prompt Studio provides real-time logs, task status updates, and the ability to pause, cancel, or modify tasks mid-execution. You can also set constraints such as:

• Maximum tokens per response.
• Timeout limits for each agent.
• Allowed file modifications (read-only, write, or append).
• Approval gates before changes are applied to your codebase.

How to Use Prompt Studio

Your First Agent Task

Let’s start with a simple task: asking an agent to generate a comment for a function. Navigate to your project directory and run:

prompt-studio run "Add a JSDoc comment to the function calculateTotal in src/utils.js"

Prompt Studio will:

1. Parse your request.
2. Read the specified file.
3. Send the context to the AI model.
4. Apply the generated comment to the file.

You will see a summary of what was changed. By default, Prompt Studio asks for confirmation before writing to disk, giving you a chance to review the output.

Running Parallel Agent Tasks

To run multiple tasks in parallel, create a JSON file called tasks.json in your project root:

[
  {
    "id": "task-1",
    "prompt": "Add error handling to the function fetchUserData in src/api.js",
    "model": "gpt-4",
    "max_tokens": 500
  },
  {
    "id": "task-2",
    "prompt": "Write unit tests for the function validateEmail in src/validators.js",
    "model": "gpt-4",
    "max_tokens": 1000
  },
  {
    "id": "task-3",
    "prompt": "Refactor the loop in src/process.js to use Array.map instead of forEach",
    "model": "claude-3-opus",
    "max_tokens": 800
  }
]

Then execute the batch:

prompt-studio batch --file tasks.json --parallel 3

The --parallel flag specifies how many agents to run simultaneously. Prompt Studio will launch all three tasks at once and display progress in real-time. You can monitor the output in the terminal or open the dashboard for a visual overview.

Using the Dashboard

For a more visual experience, launch the dashboard:

prompt-studio dashboard

Open your browser and navigate to http://localhost:3000. The dashboard displays:

• Active Tasks: A list of currently running agents with their status (running, pending, completed, or failed).
• Task Details: Click on any task to see the full prompt, model response, and any file changes made.
• Logs: A real-time log stream showing all agent activity.
• Controls: Buttons to pause, resume, or cancel individual tasks or the entire batch.

Advanced Workflow: Chaining Tasks

Prompt Studio also supports task chaining, where the output of one agent becomes the input for another. This is useful for multi-step workflows like:

1. Agent A generates a new function.
2. Agent B writes documentation for that function.
3. Agent C creates unit tests based on the documentation.

To chain tasks, use the depends_on field in your task definition:

[
  {
    "id": "generate-function",
    "prompt": "Write a function that sorts an array of objects by a given key",
    "output_file": "src/sortByKey.js"
  },
  {
    "id": "document-function",
    "prompt": "Add JSDoc comments to the file src/sortByKey.js",
    "depends_on": ["generate-function"]
  },
  {
    "id": "test-function",
    "prompt": "Write comprehensive unit tests for the functions in src/sortByKey.js",
    "depends_on": ["document-function"]
  }
]

Run the chained batch with:

prompt-studio batch --file chain-tasks.json

Prompt Studio will execute generate-function first. Once completed, document-function will start automatically, followed by test-function.

Managing Agent Permissions

To maintain control, you can restrict what agents are allowed to do. Use the --permissions flag when running tasks:

prompt-studio run "Refactor src/oldCode.js" --permissions read-write

Available permission levels:

• read-only: Agents can read files but not modify them. Useful for code review tasks.
• read-write: Agents can read and write to existing files but cannot create new files.
• full: Agents can create, read, write, and delete files. Use with caution.

You can also set global permissions in your configuration file to ensure all tasks adhere to your security policies.

Tips for Using Prompt Studio Effectively

Start Small and Iterate

If you are new to AI agent orchestration, begin with simple, single-agent tasks. Get comfortable with the CLI commands and the dashboard before scaling to parallel batches. A good starting point is asking an agent to add comments to a single file or generate a small function. Once you understand the output format and the speed of the model, gradually increase complexity.

Use Specific Prompts for Better Results

AI agents perform best when given clear, specific instructions. Instead of saying “improve this code,” try “add input validation to the function registerUser in auth.js to ensure email and password fields are not empty.” The more context you provide, the more accurate and useful the agent’s output will be. Include file paths, function names, and desired outcomes in your prompts.

Leverage the Dashboard for Monitoring

When running multiple parallel agents, the terminal can become cluttered with logs. Use the dashboard as your primary monitoring tool. It provides a clean interface for tracking progress, reviewing agent outputs, and managing tasks. You can also export logs from the dashboard for later analysis or auditing.

Set Reasonable Parallelism Limits

While Prompt Studio supports running many agents in parallel, be mindful of your API rate limits and your local machine’s resources. If you are using a cloud API like OpenAI, running 10 agents simultaneously may hit rate limits and cause delays. Start with 2–3 parallel agents and increase gradually. For local models via Ollama, your GPU or CPU memory will be the bottleneck.

Always Review Before Applying Changes

By default, Prompt Studio prompts you to review changes before writing them to disk. Do not skip this step. AI agents can make mistakes, introduce security vulnerabilities, or produce code that does not align with your project’s style. Use the diff preview feature to see exactly what will change, and only approve modifications that meet your standards.

Version Control Integration

Before running any batch of tasks, ensure your working directory is clean and committed to version control (git). Prompt Studio does not automatically create backups, so if an agent makes an unwanted change, you can revert using git. A good workflow is:

• Commit your current state.
• Run your Prompt Studio tasks.
• Review all changes.
• Commit the approved changes with a descriptive message.

Use Environment Variables for API Keys

For security, avoid hardcoding API keys in your configuration files. Instead, set them as environment variables:

export OPENAI_API_KEY="your-key-here"

Then configure Prompt Studio to read from the environment variable during the init step. This prevents accidental exposure of your keys in version control or shared configuration files.

Explore Community Workflows

Prompt Studio has an active open-source community that shares workflow templates and task configurations. Visit the official website and the project’s GitHub repository to find pre-built task definitions for common scenarios like code review, test generation, documentation, and refactoring. Adapt these templates to your own projects to save time and learn best practices.

Schedule Regular Agent Runs

For ongoing projects, consider scheduling Prompt Studio tasks to run at regular intervals using cron jobs (macOS/Linux) or Task Scheduler (Windows). For example, you could run a daily agent to scan for TODO comments and generate implementation suggestions, or a weekly agent to refactor deprecated code patterns. This turns Prompt Studio into a continuous improvement tool for your codebase.

Conclusion

Prompt Studio empowers developers to harness the full potential of AI coding agents without losing control or scalability. By combining a powerful CLI with an intuitive dashboard, it simplifies the orchestration of parallel agents, making complex workflows manageable even for beginners. Whether you are automating routine coding tasks, scaling your team’s productivity, or experimenting with multi-agent systems, Prompt Studio provides the infrastructure you need.

Start with the installation guide in this tutorial, run your first single-agent task, and then gradually explore parallel batches and chained workflows. As you become more comfortable, you will discover that Prompt Studio is not just a tool—it is a platform for reimagining how software development gets done. Visit prompt.studio to download the latest version, explore the documentation, and join the community of developers who are building the future of AI-assisted coding.