Use advanced Rovo Dev CLI features
Rovo Dev CLI is currently in beta
Rovo Dev CLI has some advanced features to help you add more context, reuse common prompts, and access tools. If you’re just getting started with Rovo Dev CLI, read the Rovo Dev CLI quickstart guide.
Memory
You can use a “memory” file to store common instructions so you don’t have to repeat them every time you use Rovo Dev. This might include coding standards, common commands, or custom instructions.
There are three places you can store your memory files, and each has different purposes:
Personal memory (global)
Your personal custom instructions that apply to all projects
Applies whenever you use Rovo Dev CLI
Stored in
~/.agent.md
Personal memory (repository)
Your personal instructions specific to a repository
Only applies in that directory
Stored in
./.agent.local.md
Team memory
Team-wide instructions shared by everyone using Rovo Dev CLI
Only applies in that directory
Stored in
./.agent.mdor subdirectories
Create a memory file automatically
You can get Rovo Dev to create a memory file based on the current codebase, including an overview of the code, directory structure, technologies and frameworks, and testing procedures. This lets Rovo Dev use this more detailed understanding of the repository when responding to your prompts.
To do this, run /memory init while in interactive mode. The file will be created in your current directory so you can edit it and add custom instructions, and will not overwrite any current memories you already have in the file.
Model Context Protocol (MCP) servers
Model Context Protocol (MCP) is a protocol that lets Rovo Dev access external data sources to use as context for its work.
Set your preferred MCP servers in ~/.rovodev/mcp.json.
An example MCP server setup (using freely available example MCP servers) looks like:
{
"mcpServers": {
"fetch": {
"command": "uvx",
"args": ["mcp-server-fetch"]
},
"Azure MCP Server": {
"command": "npx",
"args": [
"-y",
"@azure/mcp@latest",
"server",
"start"
]
},
"Framelink Figma MCP": {
"command": "npx",
"args": ["-y",
"figma-developer-mcp",
"--figma-api-key=<your API key>",
"--stdio"]
}
}
}Follow the instructions for the specific MCP servers you want to connect to. Some common examples are:
Instructions
Instructions are prompts you have saved to re-use later.
To set up and manage your instructions, edit the file .rovodev/instructions.yml.
In interactive mode, use /instructions to see your collection of saved tasks. Then, select any task from the list to run it.
Tools
You can install useful tools on your local machine and give Rovo Dev permission to use them, for example:
GitHub CLI
Code interpreter/REPL (like Python or Node.js)
Linters and formatters (black, flake8, eslint, prettier)
Test runners (pytest, unittest, jest)
Local web servers (like Flask, FastAPI, or Express)
When Rovo Dev uses a tool, it will ask you for permission first. You can choose to always allow or deny permission to use that tool to avoid being asked repeatedly.
You can change permission decisions that you have made by editing the permissions in ~/.rovodev/config.yml.
Technical information
Session files are stored in <user home dir>/.rovodev/sessions/.
Rovo Dev respects .gitignore files.
Rovo Dev uses AI models from Anthropic: Sonnet 4, Sonnet 3.7 and Sonnet 3.5v2. We use either Amazon AWS Bedrock or Google Vertex AI as a model service provider. We may switch between these models and providers, depending on capacity.
Was this helpful?