Agentic coding
Agentic coding is software development where an LLM-powered agent plans, edits, runs, and verifies code on its own using tools, not just autocomplete.
Agentic coding is the practice of letting an LLM-powered agent take multi-step actions on a codebase — reading files, editing them, running tests, checking output, and iterating — instead of returning a single completion. The agent decides what to do next based on what it observes, closing the loop between "suggest" and "do." Tools like Claude Code, Codex CLI, Qwen Code, and Aider are all agentic.
Why it matters
Traditional autocomplete tools (early Copilot, plain ChatGPT pasted into an editor) leave all the orchestration to you: you run the tests, you apply the diff, you tell it what failed. Agentic coding moves that loop inside the tool. You describe the goal — "fix the flaky auth test" — and the agent investigates, patches, runs the test, and re-patches until it passes.
This changes how developers structure work. Instead of one session per prompt, you often want several agents running in parallel on different branches or tasks. SpaceSpider exists specifically for this: it hosts a grid layout of agentic CLIs in one window so you can watch several autonomous agents work at once. See getting-started.
How it works
An agentic coding session is a loop: the agent receives a goal, proposes a tool use call (read file, run shell command, apply diff), the CLI wrapper executes it, and the result is appended to the context window as observation. The LLM reads the observation and emits the next tool call. Loop until the goal is met or the agent hits a stopping condition.
Most modern agentic CLIs add plan mode for upfront strategy, checkpoints to roll back changes, hooks to customize behavior, and sandboxing to limit blast radius. The Model Context Protocol is emerging as a shared way to plug external tools (databases, APIs, browsers) into any agent.
How it's used
Typical agentic coding tasks:
- "Migrate this file from Jest to Vitest"
- "Find and fix all N+1 queries in the orders service"
- "Write a failing test that reproduces issue #482, then fix it"
- "Refactor this component to use the new design system"
Running several of these in parallel panes — one per task, each in its own git worktree — is a common SpaceSpider pattern. See /compare/claude-code-vs-cursor for how agentic CLIs differ from editor-embedded AI.
Related terms
- Autonomous agent — the general AI agent concept
- AI pair programming — the earlier interaction model
- Subagent — delegating subtasks to other agents
- Tool use — the mechanism that enables agency
- Vibe coding — agentic coding taken to its casual extreme
Related terms
- AI pair programmingAI pair programming is a collaboration style where an LLM assistant sits alongside you, suggesting code and reviewing changes in real time as you work.
- ANSI escape codesANSI escape codes are control sequences that terminals interpret for colors, cursor movement, and screen clearing — the language of every modern CLI UI.
- Autonomous agentAn autonomous agent is an AI program that perceives, decides, and acts on its own toward a goal — the architecture behind modern coding CLIs.
- CheckpointA checkpoint is a saved snapshot of file state that lets you roll back an AI coding agent's changes to a known-good point.
- Claude CodeClaude Code is Anthropic's official command-line agent that plans, edits, runs, and verifies code across your repo using Claude models and tool use.
- CLI wrapperA CLI wrapper is a program that launches, manages, and UI-wraps other command-line tools — how apps like SpaceSpider host Claude Code and Codex CLI.