Coding Agent
A coding agent is an LLM system specialized for software-engineering tasks — reading code, editing files, running tests, executing shell commands, and iterating on results until a task is complete. The tool surface typically includes a file system, a shell, a code interpreter, and increasingly Model Context Protocol (MCP) servers for external integrations. Examples include Cursor's agent mode, Claude Code, Cline, Aider, Devin, and Replit Agent. Coding agents are distinguished from general chat assistants by both the breadth of the tool stack and the multi-step autonomous loop in which a single user request maps to many internal actions before the agent reports back.
Example
A coding agent receives "add a rate limit to the /api/login endpoint and write a test that verifies it." It locates the route handler, drafts a plan, edits the handler to integrate a rate-limit middleware, adds a new test that fires repeated requests and asserts a 429 response, runs the test suite to confirm everything passes, and presents the resulting diff for human review. The user prompted once; the agent took roughly a dozen tool actions to complete the task.
Related Resources
Bolt.new Prompting Guide (2026)
How to prompt Bolt.new — full-stack app briefs, WebContainer constraints, tech-stack specification, and the in-browser iteration loop.
Claude Code Prompting Guide (2026)
How to prompt Claude Code — Anthropic's terminal-native coding agent. CLAUDE.md files, slash commands, hooks, MCP servers, and scoped work-order prompts.
Continue.dev Prompting Guide (2026)
How to prompt Continue.dev — the open-source AI coding extension. Custom commands, context providers, model routing, and config-driven workflows.
Put this into practice
Build polished, copy-ready prompts in under 60 seconds with SurePrompts.
Try SurePrompts