Claude Code hooks let you run your own shell commands automatically at specific points in a session — before a tool runs, after a file is edited, when a session starts. They turn Claude Code from a tool you interact with into one that fits into a workflow you actually control, automatically enforcing rules you’d otherwise have to remember to do yourself.
What Are Hooks?
A hook is a shell command that Claude Code runs automatically at a defined point in its lifecycle — configured in a settings file rather than typed as a prompt. Instead of asking Claude to remember to run your linter after every edit, a hook runs it automatically, every time, with no chance of it being skipped because a conversation moved on to something else.
This matters because prompts are requests, not guarantees — Claude generally follows instructions well, but a hook is deterministic. It runs the same way every time, regardless of what’s happening in the conversation, which makes it the right tool for anything that must happen consistently rather than something that’s merely a good idea.
What You Can Hook Into
Claude Code exposes several points in its workflow where a hook can run, including before and after tool use, and at session start. Practical uses that are common by mid-2026:
- Auto-formatting. Run your formatter automatically after every file edit, so code never sits unformatted even for a moment, without needing to ask.
- Linting and type-checking. Run a linter after edits and surface errors immediately, catching issues before they compound across multiple changes.
- Blocking dangerous commands. A hook can inspect a command before it runs and block anything matching a pattern you don’t want executed — a safety net beyond just trusting the conversation.
- Logging and notifications. Log every file Claude Code touches during a session, or trigger a desktop notification when a long-running task finishes.
- Session setup. Automatically activate a virtual environment, pull the latest branch, or run a project’s setup script the moment a session starts.
How to Set One Up
Hooks are configured in Claude Code’s settings, associating an event (like “after file edit”) with a shell command to run. A minimal example — running Prettier after every edit — needs only the event name and the command itself; no special scripting language, just whatever shell command you’d normally run by hand.
Start with one hook solving one real annoyance — auto-formatting is the easiest first win — before building out a more elaborate setup. Hooks compound well once you trust the basics.
A Realistic Example
A developer working across a large team keeps hitting the same problem: code reviews flagging formatting issues Claude Code’s edits introduced. Adding a single hook that runs the project’s formatter after every file edit eliminates the problem entirely — not because Claude got better at formatting, but because the hook now guarantees it regardless.
Hooks vs Just Asking Claude
It’s worth being clear about the difference: asking Claude to “always run the linter after editing” works most of the time, because Claude generally follows instructions well. A hook works every time, because it’s not a request being interpreted — it’s a command that runs. Use hooks for anything where “most of the time” isn’t good enough: formatting standards, safety checks, and anything that needs to survive a long session without being forgotten.
Common Questions
Can hooks slow down Claude Code? A hook running after every edit does add a small delay while it executes, so it’s worth keeping hooked commands fast — a formatter or a targeted lint check is fine; a full test suite on every single edit usually isn’t, and is better reserved for a session-end or pre-commit hook instead.
Are hooks specific to one project? Hooks can be configured per-project or more broadly, depending on where the settings live — project-specific hooks are the more common setup, since formatting and linting rules usually differ between codebases.
Next Steps
Pick the one manual step you find yourself doing after almost every Claude Code session — formatting, running tests, checking a lint rule — and turn it into a hook. It’s usually a five-minute setup that removes a recurring piece of friction permanently.
Related Guides
- What is Claude Code? The AI Coding Tool Explained
- CLAUDE.md Explained: How to Configure Claude Code
- Claude Code Subagents Explained
