Dear community members!
Following our release of Rovo Dev CLI beta, we’re excited to share a comprehensive guide with prompt ideas, helpful tips, use cases and features to get you off the ground running in-product!
To learn more about Rovo Dev CLI and how to get started, see the full article here.
Features Overview
Program
|
Permission control
|
User can control Rovo Dev's permission to use any tool and file operations
|
|
File operations
|
Perform file operations on the local machine including create, read, update, and delete.
|
|
Bash commands
|
Execute shell commands with built-in permission control.
|
|
Memory support:
|
User and team can define memory files loaded into the AI agent.
|
|
MCP server support:
|
User can configure MCP servers for enhanced functionality.
|
|
Session management:
|
Automatically save and manage sessions across interactions. User can fork or resume previous sessions via an easy-to-use UI.
|
|
Task management:
|
User can save regularly used prompts into reusable tasks.
|
|
Shadow mode:
|
User can test changes without affecting the live environment.
|
|
Cost & usage tracking:
|
User can monitor the consumption of resources and expenses.
|
Intelligent Capabilities
|
Natural language understanding & interactions
|
Understand human natural language input and respond accordingly
|
|
Code analysis & reasoning:
|
Ability to analyze the entire codebase and reason with given task for complex problem solving
|
|
Generative AI:
|
Ability to generate code snippets, documentation, and solutions tailored to specific programming tasks.
|
|
Self-correction:
|
Ability to recover from errors automatically while performing a task.
|
|
Tool calling:
|
Ability to automatically use relevant tools/MCP servers to complete a task.
|
Prompt Examples
The following are common tasks with natural language prompt examples:
Planning
Coding
Testing
-
Generate synthetic data
-
"Generate 100 rows of synthetic user profile data with fields: name, email, age, location, and signup date. Output as a CSV."
-
"Create synthetic data for an e-commerce platform including orders with fields: order_id, customer_name, product, quantity, price, and order_date."
-
"Generate synthetic time series data simulating temperature sensor readings every minute for 24 hours. Include timestamp and temperature in Celsius."
-
Write tests
-
"Write unit tests in Jest for this JavaScript function that calculates user age based on birthdate"
-
"Create a test using React Testing Library to verify that my login form shows an error message when submitting an empty input"
-
"Write Pytest tests for my Flask API endpoint /users that checks GET and POST behavior"
Deploying
Modes
Interactive mode
Interactive mode is Rovo Dev’s default mode. It lets you use natural language to talk with the agent so it can answer questions, iterate on code, and problem solve with you.
Run interactive mode from the command line using
acli rovodev run
Non-interactive mode
In non-interactive mode, Rovo Dev updates files in your codebase directly without interaction. This is useful for automating repetitive tasks once you’re happy with results of an instruction.
Run non-interactive mode from the command line using
acli rovodev run <instruction>
For example, you could run
acli rovodev run "Create unit tests for all components without tests"
Commands
While in interactive mode, you can enter `/` at any time to see the list of available commands.
/models
Use models command to switch between large language models (LLMs)
/sessions
Use sessions command to switch between sessions, and view session details.
/tasks
Use tasks command to run prompts you have saved as tasks.
/memory
Use memory command to open the personal repository memory file
( ./.agent.md)
in your editor.
# <memory>
Use the above command to add a memory to the personal repository memory file ( ./.agent.md).
#! <memory>
Use the above command to remove a memory from the personal repository memory file ( ./.agent.md).
/exit
Use exit command to quit Rovo Dev.
Tips for working with Rovo Dev
-
Start small - Rovo Dev works best with small changes, about 10 or 20 lines at a time.
-
Break down your tasks - break bigger changes into smaller steps for Rovo Dev to tackle one at a time.
-
Review and iterate - treat Rovo Dev like a teammate you’re collaborating with, give feedback on the code it generates with what to change and improve.
-
Change gears - use the /sessions command to start a new session if the current approach isn’t working and you need a fresh perspective from Rovo Dev.
-
Save frequently - regularly commit changes to track your progress and avoid losing work if something goes wrong.
-
Check your work - you can ask Rovo Dev to review its own changes, compare the changes against a parent commit for linting errors, and submit draft pull requests to verify CI/CD logs.
-
Run Rovo Dev against one repository at a time - to ensure optimal performance and avoid overwhelming the system, focus on one repository rather than your entire src directory.
Remember, your use of Rovo Dev CLI must follow our Acceptable Use Policy.
Please join this community group to connect with other users and share your feedback and questions with our team!