What it is and who makes it
An Anthropic coding agent available in the terminal and other development surfaces that can read code, edit files, run commands and use integrations.
Maker or project: Anthropic. Classification: Terminal and IDE coding agent; this page treats the surrounding context, tools, permissions and verification as the relevant harness boundary.
Launch it from a project directory and provide a task with explicit scope.
The agent selects tools such as file reads, edits and shell commands.
Permission settings and hooks constrain or observe actions.
You review the diff, run tests and continue or stop the session.
Who it may fit
Pick an agent by its control surface and trust boundary, not only by its model label.
Useful scenarios
- Interactive repository work
- Multi-file changes with visible diffs and approvals
- Teams building repeatable skills, hooks and project instructions
Trade-offs and checks
- The CLI, IDE, desktop and web surfaces can have different capabilities and data paths.
- Project instructions, skills and hooks are executable workflow configuration; review them like code.
- Use read and write permissions deliberately, especially around environment files, credentials and deployment scripts.
Interactive vs headless and CI
These modes change the approval, identity and observability boundary.
Interactive terminal, IDE, desktop or web sessions are designed for human steering. For automation, use the documented non-interactive CLI or Agent SDK path, restrict permissions and treat hooks, skills and MCP as part of the execution policy.
Request-to-verification workflow
Use this mental model when debugging a surprising edit, tool call or cost.
Intent + context
Your prompt, repository instructions, selected files, history and environment become model context.
Tool + permission
The harness proposes a read, edit, shell, web, MCP or other action; policy decides whether it can run.
Result + verification
Command output, compiler errors, tests and your review return to the loop before acceptance.
Setup outline
Start with the smallest safe experiment, then follow the current official documentation for the real installation.
# Official install options change; follow the current docs. # Example workflow: cd your-project claude > First map the repository and list the files you would change. > Do not edit until I approve the plan.
Instructions, skills, MCP and hooks
These surfaces can shape context or add actions; they are part of the trust boundary.
Claude Code supports project instructions and extension surfaces such as skills, plugins, MCP servers and hooks; the exact names and capabilities are version-sensitive. Review project instruction precedence and each hook or server before enabling it.
Official sources
Use these sources for current installation, capabilities, plans, releases and security notices.
Reviewed 2026-09-26. This date is a change-tracking note, not a promise that the product remains unchanged.
Related learning
Related agent guides
Want the foundations behind the workflow?
Study AI fundamentals, prompt design, AI-assisted development, agents and MCP without a forced linear course.