Ran Boundary for a year. The security team is thrilled. The ops team is drowning.
Yes, our SSH/RDP attack surface shrank. No more direct access to prod boxes. But every "simple" task now requires a Boundary session. Developers can't just hop on a box to check logs or restart a service. They file a ticket. We get a ticket. The ticket queue has doubled.
```hcl
# Example: What used to be one ssh command is now...
# 1. Authenticate to Boundary.
# 2. Connect to the target.
# 3. Hope the session doesn't timeout mid-debug.
# 4. Log the entire session because compliance.
```
We traded a known risk for a torrent of overhead. Feels like we bought a vault to protect a broom closet. The math doesn't work for most teams.
Keep it simple
I've seen this exact trade-off happen in three different shops now. One team actually rolled back Boundary after six months because the velocity hit was real.
Have you tried scripting the auth flow? A couple bash aliases or a small CLI wrapper can cut down steps 1-2 to one command. It still logs everything, but at least you're not manually logging into the UI each time.
The timeouts are brutal though. Were you able to adjust the session TTLs, or is that locked by your security policy?
Prompt engineering is the new debugging.
Exactly. The hidden cost is never in the license, it's in the operational drag. Security gets to check a box, but the rest of us eat the productivity tax. Every ticket is a small, measurable loss that finance never sees.
You traded the "known risk" of an attack surface for the "guaranteed loss" of compounding overhead. Sounds like a bad deal to me. The math *never* works when one team's win is another team's permanent friction.
Hope you're at least charging back the time spent on tickets to the security budget.
Buyer beware