Home / Glossary / EDR (endpoint detection and response)
Glossary
What is EDR (endpoint detection and response)?
EDR (endpoint detection and response) is security software that watches laptops, desktops and servers for signs of an attack and helps the security team stop it. It records what happens on each device, such as the processes that run, the files that change and the connections that open. EDR reads behaviour, and it was not built to read what an AI agent is allowed to do.
By the AIBound team · Updated 17 November 2026
Key takeaways
- EDR records activity on each device and flags behaviour that looks like an attack.
- It gives the security team the tools to investigate and respond, such as cutting a device off from the network.
- EDR reads processes, files and behaviour, and it was not built to read what an AI agent is allowed to do.
Why EDR matters
Older antivirus tools compare files against a list of known malware. Attackers learned to work without a known bad file. They use stolen logins and the tools already built into the operating system. EDR answers this by watching behaviour over time, so a chain of odd actions stands out even when each file looks clean. Most companies run it on every managed device.
AI agents raise a different kind of question. An agent is an approved app that runs under the employee's own login, so its activity can look like the employee's normal work. The questions about an agent are which MCP servers it connects to, which skills it has loaded and what it is permitted to do. Those sit beside the questions EDR was built to answer.
How EDR works
EDR has a part on each device and a central part where the security team works.
- 1 A small sensor on each device records processes, file changes, sign-ins and network connections.
- 2 The records go to a central console, where rules and analytics look for patterns that match an attack.
- 3 When something matches, the security team gets an alert with the chain of events behind it.
- 4 The team responds from the console. It can stop a process, quarantine a file or cut the device off from the network.
EDR vs endpoint AI security
Both work on the device, and they read different things. EDR looks for an attack in progress. Endpoint AI security looks at the AI installed on the device and what that AI can reach and do. The two run side by side.
| Question | EDR | Endpoint AI security |
|---|---|---|
| What does it read? | Processes, files, sign-ins and network connections | The AI agents, MCP servers, skills and models on the device |
| What is it looking for? | Behaviour that looks like an attack | AI that is high-risk because of what it can reach and do |
| How does it act? | Stops a process, quarantines a file or isolates a device | Allows, asks about or denies an action the agent tries |
Sources
- CISA #StopRansomware Guide · cisa.gov
Sources checked 19 November 2026
Examples
EDR examples
FAQ
Frequently asked questions
What is EDR (endpoint detection and response)?
EDR is security software that runs on laptops, desktops and servers. It records what happens on each device, looks for behaviour that matches an attack and alerts the security team. The team can then investigate the chain of events and respond, for example by stopping a process or isolating the device.
What is the difference between EDR and antivirus?
Antivirus compares files against a list of known malware and blocks the ones that match. EDR watches behaviour on the device over time, so it can catch attacks that use no known bad file. EDR also keeps a record of events, which lets a security team investigate and respond after an alert.
What is the difference between EDR and XDR?
EDR watches endpoints: laptops, desktops and servers. XDR, which stands for extended detection and response, applies the same idea to more sources, such as email, identity systems, cloud services and the network. It joins those signals in one place, so an attack that crosses several systems shows up as one story.
Does EDR detect AI agents?
EDR can see that an AI agent's process is running, the same way it sees any app. It was not built to read which MCP servers and skills the agent has, or what the agent is allowed to do. AIBound reads those on the device and enforces allow, ask or deny rules inside the agent.