Ranked #5 of 710/18 · Last reviewed
Endpoint detection and response (EDR)
Endpoint agents that record processes, files and network activity on each machine and detect and stop known-malicious behaviour.
Verdict: Keep it for malware; it was not built to tell a careless agent from a careful developer.
Who it suits
Every company, as the endpoint baseline; it is rarely bought for agents.
ExamplesThe EDR already deployed on developer laptops
CostUsually already licensed per endpoint: check your current EDR contract.
Pairs with
An AASB, which judges agent actions EDR sees only as processes.
How it handles MCP connections- Discovery: Sees local MCP server processes as processes, not as MCP servers.
- Control: Can kill or block a process it judges malicious.
- Record: Process and network telemetry.
Strengths
- Usually already on every developer machine
- Records processes an agent spawns
- Stops known malware and attacker techniques
Limits
- An agent's risky actions are ordinary developer commands: rm, git push, curl
- No idea of the agent, its task or the MCP server behind a process
- Tuning it to block developer commands breaks developers' work
Scores, with reasons
Each criterion is scored 0 to 3 as defined on the criteria page.
| Criterion | Score | Why |
|---|---|---|
| Visibility | 1 | Sees processes and files, not agents, sessions or MCP servers as such. |
| Enforcement | 1 | Stops malicious patterns; a destructive but legitimate command looks normal. |
| Developer friction | 3 | Invisible to developers while it stays out of their commands. |
| Coverage of terminal and MCP actions | 1 | Sees processes a command or stdio MCP server starts, without the agent context. |
| Audit evidence | 1 | Process telemetry attributed to the user, not decisions on agent actions. |
| Deployment effort | 3 | Already deployed in most companies. |
| Total | 10/18 | Ranked #5 of 7. See the full table. |