Glossary · Coding agents
Commit
A commit is a record in a Git repository’s history that represents a project state and connects it to the project’s development. It has an identifier and associated information, including a descriptive message, helping you locate and understand the changes recorded at that point.
Commit: the explanation
In a common Git workflow, you edit files, select the content to record and then create a commit. The [Git documentation](https://git-scm.com/book/id/v2/Git-Basics-Recording-Changes-to-the-Repository) describes this preparation step. The commit represents the selected content, which is not necessarily everything currently on your computer. You can therefore record a correction separately from an unrelated documentation update. Keeping changes coherent makes the history easier to read and allows you to examine each modification with its purpose and context rather than as part of an unrelated collection.
Commits help track changes to code, instructions and configuration. You can compare project states to see what changed in an agent or identify when a feature appeared. A useful message explains the intention behind a change, while the file comparison shows what actually changed. The identifier lets you refer to a precise state when reporting a problem. This makes it easier for someone else to inspect the same code instead of guessing what you mean by the current version.
Creating a commit does not automatically publish changes to a server, pass tests or deploy software. These are separate operations, although a project may connect them through automation. A commit also provides no guarantee that its contents are correct. Additionally, removing sensitive information from the current file does not necessarily remove it from earlier commits. You need to consider the history itself when assessing what a shared repository contains, rather than looking only at the files visible in its latest state.
A concrete example
An agent helps you fix a date calculation on your website. You review the changed files and check the result. Then you create a commit dedicated to that correction, with a message explaining the problem it resolves. If something goes wrong later, you can find this project state and inspect the change without having to separate it from an unrelated update to the site’s appearance.
Why it matters
Commits make changes to an agent or tool easier to track, review and discuss. When assessing a project, their contents tell you more than their total number. Many records may be tiny or automatically generated. In your own work, coherent commits make review and reversal easier, but they do not replace meaningful tests or an appropriate backup arrangement for your files and project history.
Common misconceptions
- A commit does not automatically publish your work on the internet, because recording a local project state and sending it to a remote repository are separate operations, even when an interface offers to perform them together.
- A large number of commits does not prove rapid progress or reliable software, because their size, contents and usefulness vary, and the quality of the actual changes still needs to be examined independently of the count.
Related terms: coding agents
- GitHub stars
- GitHub stars are markers that users can add to repositories on GitHub to save them or express interest.
- fork
- A fork is a version of a software project developed separately from an existing project.
- software license
- A software license sets out the permissions and conditions that apply to using software.
- command-line interface
- A command-line interface is a way to control a program by entering text instructions, usually in a terminal.
- IDE
- An IDE is an integrated development environment, an application that brings together tools for creating and examining software.
Where this site covers it
The term appears on 3,588 pages of this site. A few of them:
- AI agent skills: how they work, how to install them, and how to build and verify them Guide
- Writing AI agent instructions: system prompts, AGENTS.md and CLAUDE.md files, and examples Guide
- Deploying AI agents in business: use cases, selection, pilots, governance, and measurement Guide
- Choosing an MCP server: trust, maintenance, permissions, and local or remote hosting Guide
- Computer use agents: how they work, available tools, reliability, and security Guide
- AI agents in October 2026: where things stand Article
- Autonomous AI agents: how much should you let them do? Article
- Malicious MCP servers and skills: check before installing Article
Go further
Frequently asked questions
Is a commit the same as saving?
Saving a file updates its contents on your storage device. Creating a commit adds selected content to Git’s history. A file can therefore be saved and contain new changes without those changes having been recorded in any commit.
Can you undo a commit?
Git offers several ways to reverse changes, including creating another commit that undoes their effects. The appropriate method depends on whether the history is shared. In collaborative work, avoid rewriting shared history without coordinating with the other contributors.
Does its message describe everything?
A commit message is written by a person or tool and may omit important details. To understand the actual change, inspect the affected files and their differences, along with any relevant checks or test results attached to that work.
Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score