Home / Glossary / AI coding agents
Glossary
What are AI coding agents?
AI coding agents are AI tools that write, test and change code on their own, instead of only suggesting it. They read a codebase, run commands and call other tools, often through MCP servers. Developers run many of them on laptops, so security teams often cannot see which ones are in use or what they can reach.
By the AIBound team · Updated 15 November 2026
Key takeaways
- An AI coding agent plans and carries out coding tasks on its own, not just a suggestion a person accepts.
- Many run on the developer's laptop, so they can sit outside gateway and network logs.
- The risk comes from what an agent can reach, such as code repositories, secrets and servers.
Why AI coding agents matter
Developers adopt AI coding agents quickly because they save hours of routine work. Each agent works with the access it is given, and that access can be wide. In one published case study, a coding agent ran with admin access across 14 code repositories at a SaaS company, and nobody had reviewed it.
How AI coding agents work
An AI coding agent turns a task into a series of steps and carries them out.
- 1 A developer gives the agent a task, such as fixing a failing build, and access to a project.
- 2 The agent reads the files, makes a plan, runs commands and edits the code.
- 3 It calls tools and servers, often through MCP, to reach tickets, databases or repositories.
- 4 The changes land in the codebase, sometimes before a person has read every step.
AI coding agents vs AI coding assistants
An AI coding assistant suggests code while a developer types, and the developer accepts or rejects it. An AI coding agent acts on its own across many files and tools, so it needs more careful limits.
| Question | AI coding agent | AI coding assistant |
|---|---|---|
| What does it do? | Plans and carries out tasks on its own | Suggests code that a person accepts |
| What can it reach? | Files, commands, tools and servers it is given | The file open in the editor |
| Where is the risk? | In what it can reach and change on its own | Mostly in the quality of the suggestion |
Sources
- Anthropic: Claude Code documentation · code.claude.com
Sources checked 19 November 2026
Examples
AI coding agent examples
FAQ
Frequently asked questions
What are AI coding agents?
AI coding agents are AI tools that write, test and change code on their own. A developer gives one a task, and it reads the project, runs commands and edits files, often calling other tools along the way. Tools such as Claude Code and Cursor can work this way.
How are AI coding agents different from AI coding assistants?
An assistant suggests code while a person types, and the person decides what to keep. An agent takes a whole task and carries it out across many files, commands and tools. Because it acts on its own, what it is allowed to reach matters much more than with an assistant.
Are AI coding agents a security risk?
They can be, depending on what they can reach. An agent with broad access to repositories, secrets or servers can change or expose far more than anyone intended. The agent itself is not the problem. The problem is an agent that security has not seen and has not set limits for.
How do you find the AI coding agents your developers use?
Look at the developer's device, because many agents run there and never touch the network. AIBound installs on employee devices in 15 minutes and finds the agents, MCP servers and skills in real time, with the person who runs each one. You then choose what to allow, ask about or deny.