Agentic Tools
Agentic tools are software tools that can perform tasks autonomously, without human intervention. They can be used in software development to automate repetitive tasks such as code generation, testing, and deployment. They can also assist developers by providing suggestions, recommendations, and insights based on the project context and the developer's preferences.
What makes a tool "agentic" is its ability to take actions — not just produce text. An agentic tool can read your files, run shell commands, call external APIs, and make decisions across multiple steps. You describe the outcome you want; the agent figures out the steps and executes them.
Popular agentic tools
There are many agentic tools available. The most popular ones are:
- Claude — a general-purpose agentic tool made by Anthropic that can perform a wide range of tasks
- Opencode — an open-source, model-agnostic terminal agent; you can plug in whichever AI model you prefer
- Codex — OpenAI's coding agent, designed for code generation and modification based on natural language descriptions
Each of these has their own client applications:
- Claude Code CLI — a command-line interface for Claude
- Opencode CLI — a command-line interface for Opencode
- Codex CLI — a command-line interface for Codex
Usually agentic CLIs can be customized to use models from different providers such as OpenAI, Anthropic, and Google. For example, you can use the Claude Code CLI to call different Anthropic models, or configure Opencode to use whichever provider you prefer.
Forms agentic tools come in
Besides the CLI, some agentic tools also have their own web applications, desktop applications, and IDE plugins:
- CLI — runs in your terminal alongside your code; highest capability and file system access
- IDE plugin — embedded in VS Code, JetBrains, etc.
- Desktop app — standalone GUI (e.g. Claude Desktop)
- Web app — browser-based, useful for quick questions
Choose whichever form best fits your workflow and preferences. For software development, the CLI is usually the most powerful option because it has direct access to your file system, shell, and tooling.
Next steps
- Setting up project context — teach the agent about your codebase
- Agentic workflow — structure your work with skills
